CN112685964B - Boiler heating surface pipe wall leakage prediction method based on convolutional neural network - Google Patents
Boiler heating surface pipe wall leakage prediction method based on convolutional neural network Download PDFInfo
- Publication number
- CN112685964B CN112685964B CN202110295637.8A CN202110295637A CN112685964B CN 112685964 B CN112685964 B CN 112685964B CN 202110295637 A CN202110295637 A CN 202110295637A CN 112685964 B CN112685964 B CN 112685964B
- Authority
- CN
- China
- Prior art keywords
- heating surface
- boiler
- neural network
- convolutional neural
- pipe wall
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to a boiler heating surface pipe wall leakage prediction method based on a convolutional neural network, which comprises the following steps: building a heating surface data matrix, mapping data to 0-255 gray values, building a heating surface temperature field in a temperature field graph building mode, taking the heating surface temperature field as input, performing classified and graded prediction on boiler heating surface pipe leakage, setting a risk grade, taking a heating surface pipe wall leakage risk grade as output, and training a CNN convolutional neural network for classified prediction; and grading the leakage risk of the tube wall of the heating surface of the boiler so as to better provide decision suggestions and support for predictive maintenance. The invention has the beneficial effects that: the invention provides a boiler heated surface pipe leakage prediction model based on a convolutional neural network; the method has important significance for detecting leakage faults and grading risks of the heated surface pipe of the boiler in the coal-fired power plant.
Description
Technical Field
The invention belongs to the technical field of reliability maintenance engineering, and particularly relates to a boiler heating surface pipe wall leakage prediction method based on a convolutional neural network.
Background
The leakage accident of the heating surface of the boiler of the thermal power plant has high probability and universality and large influence, and directly influences the completion of the receiving power and the operation target; in addition, after the denitration SCR is put into operation, if the leakage of the heating surface in the hearth is not found in time, when the leakage amount is large, the serious result of SCR catalyst poisoning can be caused after the moisture content of flue gas is improved, so that how to predict the leakage of the heating surface according to the existing DCS data becomes very significant.
At present, the research on boiler heating surface pipe leakage prediction is not much, and the method is generally a similarity-based method, namely, a model is established by using normal data, and the early warning of leakage faults is carried out based on the similarity. In addition, the fault diagnosis of the boiler heating surface pipe is also carried out from the integral angle, and the researches essentially relate monitoring data and faults to realize the fault diagnosis.
However, for fault diagnosis of a boiler heating surface pipe, especially for targeted boiler heating surface pipe wall leakage prediction, further decision suggestions are needed for fault diagnosis, and for faults, risk degree division is also needed to perform, so that predictive maintenance is performed better, and the existing research is obviously lacked in this respect; in addition, most of the existing researches are based on a data-driven method or a single measuring point, and few researches combine mathematical theory and mechanism (equipment operation principle) to construct a complete fault information description.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a boiler heating surface pipe wall leakage prediction method based on a convolutional neural network.
The boiler heating surface pipe wall leakage prediction method based on the convolutional neural network comprises the following steps:
step 1, collecting wall temperature data of a heating surface pipe wall of a coal-fired power plant boiler, and constructing a heating surface data matrix;
step 2, selecting historical data for a period of time at a certain frequency, and performing point-by-point normalization processing on a heated surface data matrix by adopting a Min-Max standardization method;
step 3, slicing the multi-dimensional data matrix after normalization processing obtained in the step 2 according to a time sequence to form a two-dimensional plane matrix; mapping the two-dimensional plane matrix to gray values of 0-255 to construct a heating surface temperature field map;
step 5, training a classification prediction model based on the CNN convolutional neural network by taking the heating surface temperature field diagram as the input of the classification prediction model based on the CNN convolutional neural network and the heating surface pipe wall leakage risk level as the output of the classification prediction model based on the CNN convolutional neural network;
step 6, with the updating of the data, reconstructing and training the classification prediction model based on the CNN convolutional neural network in the step 5 after updating a preset amount of historical data;
and 7, calling the classification prediction model based on the CNN convolutional neural network trained in the step 5, and predicting and grading the leakage risk of the heating surface pipe wall: for actually measured data, from the current moment, according to the sample taking frequency and the selected measuring points trained by the classification prediction model based on the CNN convolutional neural network, tracing m time points forwards; acquiring historical data of m time points, and taking the heated surface temperature field patterns corresponding to the historical data of the m time points as the input of the classification prediction model based on the CNN convolutional neural network by adopting a normalization processing mode for training the classification prediction model based on the CNN convolutional neural network in the steps 1 to 3 and a heated surface temperature field pattern construction method;
and 8, sending the risk prediction and risk level alarm of the heating surface leakage to the user according to the output result of the classification prediction model based on the CNN convolutional neural network in the step 7.
Preferably, step 1 specifically comprises the following steps:
step 1.1, recording that each layer of the pipe wall of the heating surface of the coal-fired power plant boiler is provided with M pipes, each pipe is provided with N rows of measuring points from top to bottom, constructing an NxM heating surface data matrix, and recording the NxM heating surface data matrix as
：
In the above formula, N is the number of rows in the heating surface data matrixM is the column number of the heating surface data matrix; mean value of temperature values of measuring points of ith measuring point plane from top to bottom of jth pipe at any layer of jth pipe on heating surface pipe wall of coal-fired power plant boiler
；
Step 1.2, recording that the heating surface pipe wall of the coal-fired power plant boiler has P layers, and establishing a heating surface data matrix for the pipes of the heating surface pipe wall of each layer of the boiler according to the step 1.1.
Preferably, step 2 performs data normalization on the temperature values in each data matrix by using the following formula:
wherein the content of the first and second substances,
which represents the value of the temperature after the normalization,
represents the measured temperature value or values and,
represents the lower threshold boundary set for the wall temperature of the heating surface of the layer in the operating specification,
represents the upper threshold boundary set for the wall temperature of the heating surface of the layer in the operating specification.
Preferably, step 3 specifically comprises the following steps:
step 3.1, collecting the pipe wall temperature of the heated surface of the coal-fired power plant boiler for enough time, constructing a P-layer boiler heated surface pipe wall temperature data matrix at each moment, and recording the time dimension as n;
step 3.2, setting the minimum threshold value of the pipe wall temperature of each layer of the heating surface in the operation rule
Corresponding gray value is set to 0, maximum threshold
The corresponding gray level is set to 255; analyzing data distribution of boiler heating surface pipe wall temperature data, and dividing nonlinear 0-255 gray values according to the data distribution;
and 3.3, constructing a temperature field diagram for the data matrix of each layer of pipeline on the wall of the heating surface of the boiler based on the gray value in the range of 0-255 corresponding to the temperature data of the wall of the heating surface of the boiler.
Preferably, the step 4 specifically comprises the following steps:
step 4.1, according to the historical working conditions actually corresponding to the N × P heating surface data matrixes which are constructed, constructing a mapping relation between a temperature field diagram and the historical working conditions in a clustering and mode matching mode; wherein N is a time dimension, P is the number of layers of the pipe wall of the heating surface of the coal-fired power plant boiler, N is the number of rows of the data matrix of the heating surface, and M is the number of columns of the data matrix of the heating surface;
4.2, dividing the historical working conditions into 1-5 risk levels, wherein 5 is the highest risk, 1 is the lowest risk, and the risk level is increased along with the increasing of the risk level; and constructing a mapping relation between the temperature field map and the risk level.
Preferably, step 5 specifically comprises the following steps:
step 5.1, grouping a plurality of temperature field graphs according to a time sequence by a fixed number m to form a temperature field graph group;
step 5.2, classifying the temperature field graph group mapped by a large amount of historical data according to 5 leakage risk levels: dividing the historical working conditions into 1-5 level risk levels; wherein, the highest risk is grade 5, the lowest risk is grade 1, and the mapping relation between the temperature field diagram and the risk grade is constructed; and obtaining a trained classification prediction model based on the CNN convolutional neural network.
The invention has the beneficial effects that: the invention provides a boiler heating surface pipe leakage prediction model based on a convolutional neural network aiming at the problem of boiler heating surface pipe wall leakage prediction, the invention utilizes boiler heating surface temperature historical data to construct a heating surface data matrix, maps the data to 0-255 gray scale values, establishes fault information description combining mechanism and mathematical theory in a way of constructing a temperature field diagram, constructs a heating surface temperature field, takes the heating surface temperature field as input, carries out classified and graded prediction on boiler heating surface pipe leakage, sets a risk grade, takes the heating surface pipe wall leakage risk grade as output, and trains a CNN convolutional neural network for classified prediction; the invention grades the leakage risk of the boiler heating surface pipe wall to provide decision suggestions and support for predictive maintenance better; the method has important significance for detecting leakage faults and grading risks of the heated surface pipe of the boiler in the coal-fired power plant.
Drawings
FIG. 1 is a flow chart of a method for predicting leakage of a tube wall of a heating surface of a boiler based on a convolutional neural network;
FIG. 2 is a diagram of prediction of tube wall leakage at the heating surface of a boiler.
Detailed Description
The present invention will be further described with reference to the following examples. The following examples are set forth merely to aid in the understanding of the invention. It should be noted that, for a person skilled in the art, several modifications can be made to the invention without departing from the principle of the invention, and these modifications and modifications also fall within the protection scope of the claims of the present invention.
According to the invention, a heating surface data matrix is constructed by using historical data of the temperature of the heating surface of the boiler, and the data is mapped to 0-255 gray values, so that a heating surface temperature field is constructed. And on the other hand, the wind grade of the leakage risk of the heating surface pipe wall is carried out based on expert knowledge. And training a classification prediction model based on a CNN convolutional neural network to perform classification prediction by taking the temperature field of the heating surface as input and the pipe wall leakage risk grade of the heating surface as output, thereby realizing the prediction and grading of the pipe wall leakage risk of the heating surface.
Example 1:
a construction process of a classification prediction model based on a CNN convolutional neural network comprises the following specific steps:
step 1, collecting wall temperature data of a heating surface pipe wall of a coal-fired power plant boiler, and constructing a heating surface data matrix;
step 1.1, recording that each layer of the pipe wall of the heating surface of the coal-fired power plant boiler is provided with M pipes, each pipe is provided with N rows of measuring points from top to bottom, constructing an NxM heating surface data matrix, and recording the NxM heating surface data matrix as
：
In the above formula, N is the number of rows of the heating surface data matrix, and M is the number of columns of the heating surface data matrix; mean value of temperature values of measuring points of ith measuring point plane from top to bottom of jth pipe at any layer of jth pipe on heating surface pipe wall of coal-fired power plant boiler
；
Step 1.2, recording that the pipe wall of the heating surface of the boiler of the coal-fired power plant has P layers, and establishing a heating surface data matrix for the pipe of the pipe wall of the heating surface of each layer of the boiler according to the step 1.1;
step 2, selecting historical data for a period of time at a certain frequency, and performing point-by-point normalization processing on a heated surface data matrix by adopting a Min-Max standardization method; step 2, for the temperature value in each data matrix, carrying out data normalization by adopting the following formula:
wherein the content of the first and second substances,
which represents the value of the temperature after the normalization,
represents the measured temperature value or values and,
represents the lower threshold boundary set for the wall temperature of the heating surface of the layer in the operating specification,
indicating an upper threshold boundary set for the wall temperature of the heated surface of the layer during the operating schedule
Step 3, slicing the multi-dimensional data matrix after normalization processing obtained in the step 2 according to a time sequence to form a two-dimensional plane matrix; mapping the two-dimensional plane matrix to gray values of 0-255 to construct a heating surface temperature field map;
step 3.1, collecting the pipe wall temperature of the heated surface of the coal-fired power plant boiler for enough time, constructing a P-layer boiler heated surface pipe wall temperature data matrix at each moment, and recording the time dimension as n;
step 3.2, setting the minimum threshold value of the pipe wall temperature of each layer of the heating surface in the operation rule
Corresponding gray value is set to 0, maximum threshold
The corresponding gray level is set to 255; analyzing data distribution of boiler heating surface pipe wall temperature data, and dividing nonlinear 0-255 gray values according to the data distribution;
3.3, constructing a temperature field diagram for the data matrix of each layer of pipeline on the wall of the heating surface of the boiler based on the gray value in the range of 0-255 corresponding to the temperature data of the wall of the heating surface of the boiler;
step 4.1, according to the historical working conditions actually corresponding to the N × P heating surface data matrixes which are constructed, constructing a mapping relation between a temperature field diagram and the historical working conditions in a clustering and mode matching mode; wherein N is a time dimension, P is the number of layers of the pipe wall of the heating surface of the coal-fired power plant boiler, N is the number of rows of the data matrix of the heating surface, and M is the number of columns of the data matrix of the heating surface;
4.2, dividing the historical working conditions into 1-5 risk levels, wherein 5 is the highest risk, 1 is the lowest risk, and the risk level is increased along with the increasing of the risk level; constructing a mapping relation between a temperature field diagram and a risk level;
step 5, training a classification prediction model based on the CNN convolutional neural network by taking the heating surface temperature field diagram as the input of the classification prediction model based on the CNN convolutional neural network and the heating surface pipe wall leakage risk level as the output of the classification prediction model based on the CNN convolutional neural network;
step 5.1, grouping a plurality of temperature field graphs according to a time sequence by a fixed number m to form a temperature field graph group;
step 5.2, classifying the temperature field graph group mapped by a large amount of historical data according to 5 leakage risk levels: dividing the historical working conditions into 1-5 level risk levels; wherein, the highest risk is grade 5, the lowest risk is grade 1, and the mapping relation between the temperature field diagram and the risk grade is constructed; and obtaining a trained classification prediction model based on the CNN convolutional neural network.
Example 2:
in this embodiment, a trained classification prediction model based on a CNN convolutional neural network is used to predict the boiler heating surface tube wall leakage, and on the basis of embodiment 1, the present embodiment further includes the following steps:
step 6, with the updating of the data, reconstructing and training the classification prediction model based on the CNN convolutional neural network in the step 5 after updating a preset amount of historical data;
and 7, calling the classification prediction model based on the CNN convolutional neural network trained in the step 5, and predicting and grading the leakage risk of the heating surface pipe wall: for actually measured data, from the current moment, according to the sample taking frequency and the selected measuring points trained by the classification prediction model based on the CNN convolutional neural network, tracing m time points forwards; acquiring historical data as input of a classification prediction model based on a CNN (convolutional neural network), and mapping measured data into temperature field data as input of the classification prediction model based on the CNN in steps 1 to 3 by adopting a normalization processing mode for training the classification prediction model based on the CNN and a heating surface temperature field map construction method; calling a trained classification prediction model based on a CNN convolutional neural network to predict and grade the leakage risk of the heating surface pipe wall;
and 8, sending the risk prediction and risk level alarm of the heating surface leakage to the user according to the output result of the classification prediction model based on the CNN convolutional neural network in the step 7.
Example 3:
by combining the embodiment 1 and the embodiment 2, the structure of the heating surface of the water wall of the boiler in the embodiment is formed by the vertical water wall, the operating temperature range of the water wall of the boiler is 310-480 ℃ in the design working condition, a large number of wall temperature sensors are arranged on the heating surface, and the overtemperature alarm value of the wall temperature of the water wall is 460 ℃. In the embodiment, when the boiler operates at load 550MW, the overtemperature alarm is given to the heating surface of the water-cooled wall, and the heating surface is damaged due to continuous overtemperature to cause leakage fault, so that the output of the boiler is limited, and finally the shutdown fault of a unit is caused.
As shown in fig. 1, a method for predicting boiler heating surface tube wall leakage based on a convolutional neural network specifically executes the following processes:
step 1, collecting wall temperature data of a heating surface pipe wall of a coal-fired power plant boiler, and constructing a heating surface data matrix;
step 1.1, recording that each layer of the pipe wall of the heating surface of the coal-fired power plant boiler is provided with M pipes, each pipe is provided with N rows of measuring points from top to bottom, constructing an NxM heating surface data matrix, and recording the NxM heating surface data matrix as
：
In the above formula, N is the number of rows of the heating surface data matrix, and M is the number of columns of the heating surface data matrix; j-th pipe on any layer of heating surface pipe wall of coal-fired power plant boilerThe average value of the temperature values of all measuring points of the ith measuring point plane of the tube from top to bottom is
(ii) a The heating surface selected in this embodiment is composed of 30 tubes, each tube has 30 rows of measuring points from top to bottom, the heating surface data matrix is 30 × 30, the heating surface data matrix is divided, 3 × 3 is taken as a calculation surface, and the mean value of each block is composed into a 10 × 10 heating surface data matrix, so that the finally obtained heating surface data matrix X is a 10 × 10 data matrix, and the training data of X at a certain time is as shown in table 1 below.
TABLE 1 training data Table for heating surface data matrix X at a certain time
| 408.52 | 409.2 | 414.2 | 416.6 | 418.3 | 404.6 | 411.1 | 413.7 | 413.9 | 407.2 |
| 406.9 | 406.0 | 410.1 | 413.5 | 418.5 | 406.0 | 415.05 | 416.1 | 416.0 | 408.6 |
| 408.7 | 406.1 | 409.9 | 414.6 | 420.0 | 408 | 416.5 | 420.0 | 420.6 | 412.1 |
| 410.9 | 411.2 | 415.5 | 420.3 | 422.0 | 408.6 | 416.2 | 425.0 | 425.4 | 414.3 |
| 401.4 | 403.5 | 405.8 | 409.8 | 411.52 | 398.56 | 406.27 | 414.93 | 416.07 | 405.1 |
| 401.2 | 402.1 | 404.4 | 409.3 | 413.7 | 400.6 | 409.9 | 418.0 | 418.8 | 407.0 |
| 400.3 | 399.9 | 401.3 | 406.3 | 411.7 | 399.7 | 409.0 | 415.6 | 416.5 | 406.1 |
| 399.1 | 398.3 | 399.7 | 404.4 | 410.1 | 398.6 | 408.1 | 413.4 | 414.7 | 405.3 |
| 398.6 | 398.2 | 399.2 | 402.7 | 408.8 | 397.5 | 407.5 | 411.8 | 413.8 | 403.8 |
| 397.2 | 396.2 | 396.8 | 400.1 | 406.74 | 396.3 | 405.5 | 409.1 | 410.5 | 401.6 |
Step 1.2, recording that the pipe wall of the heating surface of the boiler of the coal-fired power plant has P layers, and establishing a heating surface data matrix for the pipe of the pipe wall of the heating surface of each layer of the boiler according to the step 1.1; in the present embodiment, the P parameter is set to 1.
Step 2, selecting historical data for a period of time at a certain frequency, and performing point-by-point normalization processing on a heated surface data matrix by adopting a Min-Max standardization method; for the temperature values in each data matrix, the data normalization is performed by using the following formula:
wherein the content of the first and second substances,
which represents the value of the temperature after the normalization,
represents the measured temperature value or values and,
represents the lower threshold boundary set for the wall temperature of the heating surface of the layer in the operating specification,
represents the upper threshold value boundary set for the tube wall temperature of the heating surface of the layer in the operation specification, and the embodiment is normalized
The value range of (1) is (0); normalized temperature value of training data at a certain time
As shown in table 2 below;
TABLE 2 temperature table after training data normalization at a certain time
| 0.42 | 0.43 | 0.44 | 0.45 | 0.45 | 0.45 | 0.45 | 0.45 | 0.46 | 0.45 |
| 0.38 | 0.39 | 0.42 | 0.43 | 0.43 | 0.44 | 0.44 | 0.44 | 0.45 | 0.43 |
| 0.36 | 0.37 | 0.39 | 0.42 | 0.42 | 0.44 | 0.43 | 0.43 | 0.45 | 0.41 |
| 0.34 | 0.35 | 0.36 | 0.40 | 0.40 | 0.43 | 0.43 | 0.95 | 0.43 | 0.41 |
| 0.33 | 0.33 | 0.35 | 0.39 | 0.39 | 0.42 | 0.41 | 0.42 | 0.43 | 0.38 |
| 0.29 | 0.31 | 0.33 | 0.37 | 0.38 | 0.40 | 0.40 | 0.41 | 0.42 | 0.37 |
| 0.28 | 0.29 | 0.29 | 0.33 | 0.34 | 0.36 | 0.38 | 0.40 | 0.39 | 0.34 |
| 0.27 | 0.28 | 0.28 | 0.31 | 0.32 | 0.34 | 0.35 | 0.35 | 0.36 | 0.34 |
| 0.28 | 0.27 | 0.28 | 0.30 | 0.31 | 0.31 | 0.32 | 0.33 | 0.34 | 0.33 |
| 0.27 | 0.26 | 0.27 | 0.28 | 0.29 | 0.30 | 0.31 | 0.32 | 0.32 | 0.32 |
Step 3, slicing the multi-dimensional data matrix after normalization processing obtained in the step 2 according to a time sequence to form a two-dimensional plane matrix; mapping the two-dimensional plane matrix to gray values of 0-255 to construct a heating surface temperature field map; step 3.1, collecting the pipe wall temperature of the heating surface of the coal-fired power plant boiler for enough time, and constructing a P-layer boiler heating surface pipe wall temperature data matrix at each moment, wherein the sampling frequency is specified according to actual business requirements, generally 1 minute or 1 second, and the time dimension is n; in the embodiment, the wall temperature change of the heating surface of the boiler is stable, and the sampling frequency is set to be 1 minute;
step 3.2, setting the minimum threshold value of the pipe wall temperature of each layer of the heating surface in the operation rule
Corresponding gray value is set to 0, maximum threshold
The corresponding gray level is set to 255; analyzing data distribution of boiler heating surface pipe wall temperature data, and dividing nonlinear 0-255 gray values according to the data distribution; according to historical data of wall temperature of a water-cooled wall of a boiler, the wall temperature distribution is generally normal, 310-360 ℃ is mapped to 0-50 gray values, 361-430 ℃ is mapped to 51-200 gray values, and 431-480 ℃ is mapped to 201-255 gray values;
3.3, constructing a temperature field diagram for the data matrix of each layer of pipeline on the wall of the heating surface of the boiler based on the gray value in the range of 0-255 corresponding to the temperature data of the wall of the heating surface of the boiler;
step 4.1, according to the historical working conditions actually corresponding to the N × P heating surface data matrixes which are constructed, constructing a mapping relation between a temperature field diagram and the historical working conditions in a clustering and mode matching mode; wherein N is a time dimension, P is the number of layers of the pipe wall of the heating surface of the coal-fired power plant boiler, N is the number of rows of the data matrix of the heating surface, and M is the number of columns of the data matrix of the heating surface;
4.2, dividing the historical working conditions into 1-5 risk levels by combining knowledge in the field, wherein the 5 level is the highest risk, the 1 level is the lowest risk, and the risk level is increased along with the increment of the risk level; constructing a mapping relation between a temperature field diagram and a risk level;
step 5, training a classification prediction model based on the CNN convolutional neural network by taking the heating surface temperature field diagram as the input of the classification prediction model based on the CNN convolutional neural network and the heating surface pipe wall leakage risk level as the output of the classification prediction model based on the CNN convolutional neural network;
step 5.1, grouping a plurality of temperature field graphs according to a time sequence by a fixed number m to form a temperature field graph group;
step 5.2, classifying the temperature field graph group mapped by a large amount of historical data according to 5 leakage risk levels: dividing the historical working conditions into 1-5 level risk levels through expert knowledge; wherein, the highest risk is grade 5, the lowest risk is grade 1, and the mapping relation between the temperature field diagram and the risk grade is constructed; this embodiment ranks the risk as follows: the gray values of 0-50 are 1-level risk, the gray values of 51-100 are 2-level risk, the gray values of 101-50 are 3-level risk, the gray values of 151-200 are 4-level risk, and the gray values of 201-255 are 5-level risk; obtaining a trained classification prediction model based on a CNN convolutional neural network;
step 6, with the updating of the data, reconstructing and training the classification prediction model based on the CNN convolutional neural network in the step 5 after a preset amount of historical data is updated so as to obtain a more accurate prediction result;
and 7, calling the classification prediction model based on the CNN convolutional neural network trained in the step 5, and predicting and grading the leakage risk of the heating surface pipe wall: for actually measured data, from the current moment, according to the sample taking frequency and the selected measuring points trained by the classification prediction model based on the CNN convolutional neural network, tracing m time points forwards; acquiring historical data as input of a classification prediction model based on a CNN (CNN convolutional neural network) (m is set according to actual service requirements), adopting a normalization processing mode for training the classification prediction model based on the CNN convolutional neural network in the steps 1 to 3 and a heating surface temperature field map construction method, and mapping measured data into temperature field data (a data normalization and temperature field mapping method) as input of the classification prediction model based on the CNN convolutional neural network; calling a trained classification prediction model based on a CNN convolutional neural network to predict and grade the leakage risk of the heating surface pipe wall;
step 8, according to the output result of the classification prediction model based on the CNN convolutional neural network in the step 7, heating surface leakage risk prediction and risk level alarm are sent to a user;
m of the fixed number m of sheets in fig. 1 is equivalent to m of m times.
The prediction result of the boiler heating surface tube wall leakage in this embodiment is shown in fig. 2, the classification prediction model based on the CNN convolutional neural network realizes prediction and grading of the tube wall leakage risk, and the positions (8, 4) in the graph output the leakage position and risk level, the gray value 243, and the risk level 5. The leakage fault of the heating surface of the water wall of the boiler is intuitively predicted in a graphical mode, the leakage position and the risk level are positioned, and according to the leakage prediction, operators can be guided to change operation, the operation condition of the boiler is adjusted, the heat load of the heating surface of the boiler is improved, and the operation safety of the heating surface of the boiler is improved. Meanwhile, according to the prediction result, the unit can be reasonably and safely overhauled, spare parts can be purchased and overhauled, and the inventory and loss of the spare parts are reduced.
Claims (4)
1. A boiler heating surface pipe wall leakage prediction method based on a convolutional neural network is characterized by comprising the following steps:
step 1, collecting wall temperature data of a heating surface pipe wall of a coal-fired power plant boiler, and constructing a heating surface data matrix;
step 2, selecting historical data for a period of time at a fixed frequency, and performing point-by-point normalization processing on a heated surface data matrix by adopting a Min-Max standardization method;
step 3, slicing the multi-dimensional data matrix after normalization processing obtained in the step 2 according to a time sequence to form a two-dimensional plane matrix; mapping the two-dimensional plane matrix to gray values of 0-255 to construct a heating surface temperature field map;
step 3.1, acquiring the pipe wall temperature of the heating surface of the coal-fired power plant boiler, constructing a P-layer boiler heating surface pipe wall temperature data matrix at each moment, and recording a time dimension as n;
step 3.2, setting the minimum threshold value of the pipe wall temperature of each layer of the heating surface in the operation rule
Corresponding gray value is set to 0, maximum threshold
The corresponding gray level is set to 255; analyzing data distribution of boiler heating surface pipe wall temperature data, and dividing nonlinear 0-255 gray values according to the data distribution;
3.3, constructing a temperature field diagram for the data matrix of each layer of pipeline on the wall of the heating surface of the boiler based on the gray value in the range of 0-255 corresponding to the temperature data of the wall of the heating surface of the boiler;
step 4, based on the heating surface data matrix constructed in the step 1 and a heating surface temperature field diagram corresponding to the heating surface data matrix, defining the leakage risk level of the heating surface pipe wall according to historical working conditions;
step 4.1, according to the historical working conditions actually corresponding to the N × P heating surface data matrixes which are constructed, constructing a mapping relation between a temperature field diagram and the historical working conditions in a clustering and mode matching mode; wherein N is a time dimension, P is the number of layers of the pipe wall of the heating surface of the coal-fired power plant boiler, N is the number of rows of the data matrix of the heating surface, and M is the number of columns of the data matrix of the heating surface;
4.2, dividing the historical working conditions into 1-5 risk levels, wherein 5 is the highest risk, 1 is the lowest risk, and the risk level is increased along with the increasing of the risk level; constructing a mapping relation between a temperature field diagram and a risk level;
step 5, training a classification prediction model based on the CNN convolutional neural network by taking the heating surface temperature field diagram as the input of the classification prediction model based on the CNN convolutional neural network and the heating surface pipe wall leakage risk level as the output of the classification prediction model based on the CNN convolutional neural network;
step 6, with the updating of the data, reconstructing and training the classification prediction model based on the CNN convolutional neural network in the step 5 after updating a preset amount of historical data;
and 7, calling the classification prediction model based on the CNN convolutional neural network trained in the step 5, and predicting and grading the leakage risk of the heating surface pipe wall: for actually measured data, from the current moment, according to the sample taking frequency and the selected measuring points trained by the classification prediction model based on the CNN convolutional neural network, tracing m time points forwards; acquiring historical data of m time points, and taking the heated surface temperature field patterns corresponding to the historical data of the m time points as the input of the classification prediction model based on the CNN convolutional neural network trained in the step 5 by adopting the normalization processing mode of the classification prediction model based on the CNN convolutional neural network trained in the steps 1 to 3 and the construction method of the heated surface temperature field patterns;
and 8, sending the risk prediction and risk level alarm of the heating surface leakage to the user according to the output result of the classification prediction model based on the CNN convolutional neural network in the step 7.
2. The method for predicting the boiler heating surface pipe wall leakage based on the convolutional neural network as claimed in claim 1, wherein the step 1 specifically comprises the following steps:
step 1.1, recording that each layer of the pipe wall of the heating surface of the coal-fired power plant boiler is provided with M pipes, each pipe is provided with N rows of measuring points from top to bottom, constructing an NxM heating surface data matrix, and recording the NxM heating surface data matrix as
：
In the above formula, N is the number of rows of the heating surface data matrix, and M is the number of columns of the heating surface data matrix; slave of j-th pipe on any layer of pipe wall of heating surface of coal-fired power plant boilerThe average value of the temperature values of all measuring points on the plane of the upper measuring point to the lower ith measuring point is
，
；
Step 1.2, recording that the heating surface pipe wall of the coal-fired power plant boiler has P layers, and establishing a heating surface data matrix for the pipes of the heating surface pipe wall of each layer of the boiler according to the step 1.1.
3. The method for predicting the boiler heating surface pipe wall leakage based on the convolutional neural network as claimed in claim 1, wherein the step 2 performs data normalization on the temperature values in each data matrix by using the following formula:
wherein the content of the first and second substances,
which represents the value of the temperature after the normalization,
represents the measured temperature value or values and,
indicating the lower threshold boundary set for each floor heating surface wall temperature in the operating protocol,
representing the upper threshold boundary set for each floor heating surface wall temperature in the operating protocol.
4. The method for predicting the boiler heating surface tube wall leakage based on the convolutional neural network as claimed in claim 1, wherein the step 5 specifically comprises the following steps:
step 5.1, grouping a plurality of temperature field graphs according to a time sequence by a fixed number m to form a temperature field graph group;
step 5.2, classifying the temperature field graph group mapped by a large amount of historical data according to 5 leakage risk levels: dividing the historical working conditions into 1-5 level risk levels; wherein, the highest risk is grade 5, the lowest risk is grade 1, and the mapping relation between the temperature field diagram and the risk grade is constructed; and obtaining a trained classification prediction model based on the CNN convolutional neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110295637.8A CN112685964B (en) | 2021-03-19 | 2021-03-19 | Boiler heating surface pipe wall leakage prediction method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110295637.8A CN112685964B (en) | 2021-03-19 | 2021-03-19 | Boiler heating surface pipe wall leakage prediction method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112685964A CN112685964A (en) | 2021-04-20 |
| CN112685964B true CN112685964B (en) | 2021-06-08 |
Family
ID=75455656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110295637.8A Active CN112685964B (en) | 2021-03-19 | 2021-03-19 | Boiler heating surface pipe wall leakage prediction method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112685964B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113836821B (en) * | 2021-10-26 | 2023-11-28 | 华电莱州发电有限公司 | Online prediction method for boiler water wall cracking |
| CN114001360B (en) * | 2021-12-31 | 2022-04-26 | 南京沃谱瑞环境研究院有限公司 | Burning rotary kiln intelligent control method and system based on Internet of things |
| CN114484409B (en) * | 2022-02-22 | 2023-04-07 | 北京博数智源人工智能科技有限公司 | Early warning method and device for furnace tube leakage accident of thermal power plant |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090141966A1 (en) * | 2007-11-30 | 2009-06-04 | Microsoft Corporation | Interactive geo-positioning of imagery |
| CN108845075A (en) * | 2018-04-25 | 2018-11-20 | 南京农业大学 | Compost maturity real-time predicting method based on deep learning network |
| CN111582472A (en) * | 2020-04-17 | 2020-08-25 | 广西电网有限责任公司电力科学研究院 | Water spray attemperator water spray adjusting method and device based on neural network model |
-
2021
- 2021-03-19 CN CN202110295637.8A patent/CN112685964B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090141966A1 (en) * | 2007-11-30 | 2009-06-04 | Microsoft Corporation | Interactive geo-positioning of imagery |
| CN108845075A (en) * | 2018-04-25 | 2018-11-20 | 南京农业大学 | Compost maturity real-time predicting method based on deep learning network |
| CN111582472A (en) * | 2020-04-17 | 2020-08-25 | 广西电网有限责任公司电力科学研究院 | Water spray attemperator water spray adjusting method and device based on neural network model |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于卷积神经网络和纵横交叉优化算法的电缆隧道温度异常识别方法;孟安波等;《现代信息科技》;20190125;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112685964A (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112685964B (en) | Boiler heating surface pipe wall leakage prediction method based on convolutional neural network | |
| CN117014472B (en) | Cloud side end cooperation-based intelligent power plant equipment inspection method and system | |
| CN108052007B (en) | Thermal power generating unit operation optimization method, device and equipment and computer storage medium | |
| CN116308965B (en) | Intelligent gas underground gas pipe network safety management method, internet of things system and device | |
| CN109540212A (en) | A kind of cable shaft status assessment early warning system and method based on big data | |
| CN112990546A (en) | Chemical plant power transformer fault prediction method based on particle swarm and neural network | |
| CN112746934A (en) | Method for diagnosing fan fault through self-association neural network | |
| CN110737975B (en) | Wind power plant wind speed and power prediction and abnormality correction method based on empirical mode decomposition and autoregressive model | |
| CN117544998B (en) | Wireless network speed intelligent evaluation system based on artificial intelligence | |
| CN116862081A (en) | Operation and maintenance method and system for pollution treatment equipment | |
| CN113486950B (en) | Intelligent pipe network water leakage detection method and system | |
| CN110096723B (en) | High-voltage switch cabinet insulation state analysis method based on operation and maintenance detection big data | |
| CN117705121A (en) | Multi-dimensional inspection system for power plant | |
| CN109978271A (en) | A kind of thermal power plant's booting policy optimization method and system based on haze prediction | |
| CN116429269A (en) | Infrared intelligent analysis system for ethylene cracking furnace tube | |
| CN112800672B (en) | Evaluation method, system, medium and electronic equipment for boiler fouling coefficient | |
| CN112884170A (en) | Predictive intelligent operation and maintenance system and method for comprehensive pipe gallery | |
| CN116257963B (en) | Power plant pipeline state parameter confirmation method and device | |
| CN113836821B (en) | Online prediction method for boiler water wall cracking | |
| Gad et al. | A wind turbine fault identification using machine learning approach based on pigeon inspired optimizer | |
| CN111476383A (en) | Pump station unit state maintenance dynamic decision method | |
| CN118149606B (en) | Silane batch furnace and heating control method thereof | |
| CN117407967B (en) | Carbon emission monitoring method of integrated house | |
| CN117633589A (en) | Method and system for monitoring moving ring | |
| CN117152355B (en) | Visual supervision system based on factory data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220824 Address after: Room 307, No. 32, Gaoji Street, Xihu District, Hangzhou City, Zhejiang Province, 310002 Patentee after: Zhejiang Zheneng Digital Technology Co.,Ltd. Patentee after: ZHEJIANG ENERGY R & D INSTITUTE Co.,Ltd. Address before: 5 / F, building 1, No. 2159-1, yuhangtang Road, Wuchang Street, Yuhang District, Hangzhou City, Zhejiang Province Patentee before: ZHEJIANG ENERGY R & D INSTITUTE Co.,Ltd. |
|
| TR01 | Transfer of patent right |