CN109934417B - Boiler coking early warning method based on convolutional neural network - Google Patents
Boiler coking early warning method based on convolutional neural network Download PDFInfo
- Publication number
- CN109934417B CN109934417B CN201910232761.2A CN201910232761A CN109934417B CN 109934417 B CN109934417 B CN 109934417B CN 201910232761 A CN201910232761 A CN 201910232761A CN 109934417 B CN109934417 B CN 109934417B
- Authority
- CN
- China
- Prior art keywords
- layer
- coking
- convolution
- boiler
- neural network
- 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
- 238000004939 coking Methods 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 25
- 238000005070 sampling Methods 0.000 claims abstract description 17
- 238000003062 neural network model Methods 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims description 45
- 230000006870 function Effects 0.000 claims description 9
- 230000001502 supplementing effect Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010438 heat treatment Methods 0.000 abstract description 9
- 210000002569 neuron Anatomy 0.000 description 8
- 239000000571 coke Substances 0.000 description 7
- 239000003245 coal Substances 0.000 description 2
- 238000005235 decoking Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a boiler coking early warning method based on a convolutional neural network, which comprises the following steps: 1) Acquiring data information of coking or non-coking in the boiler; 2) Selecting temperature data of a plurality of measuring points which are coked or not coked in the same time period from the data information in the step 1); 3) Constructing a coked or non-coked convolutional neural network model, which comprises the following steps: an input layer, a convolution layer, a down-sampling layer, a full-link layer and an output layer; 4) Randomly selecting a plurality of measuring point temperature data in the same time period from a boiler data source acquired in real time, inputting the coking or non-coking convolution neural network model in the step 3), obtaining the image characteristics of the measuring point temperature data, and judging the coking or non-coking. The invention provides an early warning scheme for the coking of the boiler, can accurately predict the coking condition of the heating surface of the boiler and provides scientific guidance basis for judging the coking of the heating surface of the boiler.
Description
Technical Field
The invention relates to the technical field of operation optimization of boilers in thermal power plants.
Background
The boiler coking refers to the phenomenon that tar and dust are attached to the surface of a heating surface and gradually develop to form coke block coverage due to fuels or combustion modes in the running process of a boiler. Coking of the heating surface of the boiler is mainly concentrated on the heating surfaces such as a vertical water-cooled wall and a spiral water-cooled wall, when the coking of the heating surface is serious, parameters such as unit load, desuperheating water flow, coal mill operation mode and the like can be changed to different degrees, the service life of the heating surface of the boiler is reduced, four pipes are leaked, the unit is not stopped, and huge economic loss is caused to a power plant. At present, aiming at the phenomenon of boiler coking, a thermal power plant cannot obtain effective data of boiler coking or non-coking, and the coke interference is carried out or a decoking agent is sprayed to carry out active decoking by regularly switching the operation mode of a coal mill according to the working experience.
Disclosure of Invention
The invention aims to provide a boiler coking early warning method based on a convolutional neural network.
Based on the purpose, the invention mainly adopts the following technical scheme:
the boiler coking early warning method based on the convolutional neural network comprises the following steps:
1) Acquiring data information of coking or non-coking in the boiler;
2) Selecting temperature data of a plurality of measuring points which are coked or not coked in the same time period from the data information in the step 1);
3) Constructing a coked or non-coked convolutional neural network model, which comprises the following steps: an input layer, a convolutional layer, a downsampling layer, a full-link layer, and an output layer,
the method steps of the input layer are as follows:
a. respectively supplementing the temperature data of the plurality of measuring points which are coked or not coked in the step 2) into an a x a data matrix, and supplementing the insufficient matrix elements with 0;
b. normalizing each matrix element in the data matrix in the step a to obtain coking or non-coking analog image characteristics,
the normalized calculation formula is:
wherein: i, j are the indexes of the rows and columns of the data matrix in step a, x i，j The matrix elements for the i, j positions after normalization, vi, j to normalize the matrix elements of the previous i, j positions,
to normalize the smallest matrix element in the first j columns,
is the maximum matrix element in the normalized front j columns;
4) Randomly selecting a plurality of measuring point temperature data of the same time period from a boiler data source acquired in real time, inputting the data into the coking or non-coking convolution neural network model in the step 3) to obtain the image characteristics of the measuring point temperature data, and judging the image characteristics to be in accordance with the coking image characteristics in the convolution neural network model, wherein the data are coking; and if the image characteristics which accord with the non-coking in the convolutional neural network model are the non-coking, the image characteristics are the non-coking.
In step 3), convolutional layer: performing convolution operation on the analog image characteristics in the step b to obtain a characteristic diagram of the convolution layer;
down-sampling layer: down-sampling the feature map of the convolutional layer to obtain a feature map of the down-sampled layer;
full connection layer: flattening the feature map of the down-sampling layer to obtain a one-dimensional vector;
an output layer: and (4) performing probability output by adopting a softmax function, wherein all neurons connected with the full connection layer are contained.
In step 3), the convolutional layer is divided into a convolutional layer C1 and a convolutional layer C2, and the downsampling layer is divided into a downsampling layer S1 and a downsampling layer S2, and the method specifically comprises the following steps:
the convolutional layer C1: performing convolution operation on the analog image characteristics in the step b to obtain a characteristic diagram of the convolution layer C1;
downsampling layer S1: downsampling the feature map of the convolutional layer C1 to obtain a feature map of a downsampled layer S1;
and (3) a convolutional layer C2: performing convolution operation on the feature map of the downsampling layer S1 to obtain a feature map of a convolution layer C2;
down-sampling layer S2: the feature map of the convolutional layer C2 is downsampled to obtain the feature map of the downsampled layer S2.
In the step 2), 123 measured point temperature data in the same time period are taken, and in the step a of inputting the measured point temperature data into the layer, the size of a data matrix is 12 multiplied by 12; convolution kernels of convolution layers C1 and C2 have a mean μ =0 and a variance σ 2 Normalized positive distribution of = 1; the convolution layer C1 has 3 convolution kernels, and the number of the convolution kernels is 3 multiplied by 3; there are 5 convolution kernels in convolution layer C2, and the number of convolution kernels is 3 × 3.
The convolutional layer C1, the convolutional layer C2 and the full connection layer all adopt relu activation functions.
In step 2), 123 measured point temperature data in the same time period are taken, and in the step a of inputting the measured point temperature data into the layer, the size of the data matrix is 12 multiplied by 12.
Compared with the prior art, the invention has the following beneficial effects:
the method comprises the steps of utilizing a boiler measuring point real-time database to preprocess data by adopting an image processing idea, establishing a convolutional neural network model, utilizing an image recognition algorithm of the convolutional neural network to predict and classify the real-time data, judging whether the monitored real-time data is data in coking, providing an early warning scheme for boiler coking, accurately predicting the coking condition of a heating surface of the boiler, and providing a scientific guiding basis for judging the coking of the heating surface of the boiler.
Drawings
FIG. 1 is a block flow diagram of the present invention;
FIG. 2 is an exemplary illustration of a downsampled layer operation max pooling method;
FIG. 3 is a coking gray scale signature of wall temperature data;
FIG. 4 is a non-coking grayscale map of wall temperature data.
Detailed Description
Examples
The boiler coking early warning method based on the convolutional neural network comprises the following steps:
1) Acquiring data information of coking or non-coking in the boiler;
2) Selecting temperature data of a plurality of measuring points which are coked or not coked in the same time period from the data information in the step 1);
3) Constructing a coked or non-coked convolutional neural network model, which comprises the following steps: an input layer, a convolutional layer, a downsampling layer, a full-link layer, and an output layer,
the method steps of the input layer are as follows:
a. respectively supplementing the temperature data of the 123 measuring points which are coked or not coked in the step 2) into a 12 x 12 data matrix, and supplementing the insufficient matrix elements with 0;
b. normalizing each matrix element in the data matrix in the step a (min-max normalization) to obtain coking or non-coking analog image characteristics,
the normalized calculation formula is:
wherein: i, j are the indexes of the rows and columns of the data matrix in step a, x i，j Is the matrix element, v, of the i, j position after normalization i，j To normalize the matrix elements of the previous i, j positions,
to normalize the smallest matrix element in the first j columns,
is the maximum matrix element in the first j columns of the normalization;
4) Randomly selecting a plurality of measuring point temperature data of the same time period from a boiler data source acquired in real time, inputting the data into the coking or non-coking convolution neural network model in the step 3) to obtain the image characteristics of the measuring point temperature data, and judging the image characteristics to be in accordance with the coking image characteristics in the convolution neural network model, wherein the data are coking; and if the image characteristics which accord with the non-coking in the convolutional neural network model are the non-coking, the image characteristics are the non-coking.
Specifically, in step 3), the convolutional layer: c, comparing the analog image characteristics in the step b with the characteristics of satisfying mean value mu =0 and variance sigma 2 Carrying out convolution operation on convolution kernels in the standard normal distribution of =1 to obtain a characteristic diagram of the convolution layer;
down-sampling layer: down-sampling the feature map of the convolutional layer to obtain the feature map of the down-sampled layer;
full connection layer: performing Flatten (flattening) on the feature map of the down-sampling layer to obtain a one-dimensional vector;
an output layer: and (4) adopting a softmax function to output the probability, wherein all neurons connected with the full connection layer are contained.
Specifically, in step 3), the convolutional layer is divided into convolutional layer C1 and convolutional layer C2, the convolutional kernels of convolutional layer C1 and convolutional layer C2 are mean μ =0, and variance σ is 2 Normalized positive for =1Distributing; the down-sampling layer is divided into a down-sampling layer S1 and a down-sampling layer S2, and the specific steps are as follows:
the convolutional layer C1: c, performing convolution operation on the analog image features in the step b to obtain a feature map of the convolution layer C1, wherein the number of convolution kernels is 3, and the number of the convolution kernels is 3 multiplied by 3;
downsampling layer S1: downsampling the feature map of the convolutional layer C1 to obtain a feature map of a downsampled layer S1;
convolutional layer C2: 5 convolution kernels are provided, the number of the convolution kernels is 3 x 3, and the feature map of the downsampled layer S1 is subjected to convolution operation to obtain a feature map of a convolution layer C2;
down-sampling layer S2: the feature map of the convolutional layer C2 is downsampled to obtain the feature map of the downsampled layer S2.
The convolutional layer C1, the convolutional layer C2 and the full connection layer all adopt relu activation functions.
Examples are as follows:
the boiler coking early warning method based on the convolutional neural network has the flow shown in figure 1 and comprises the following steps:
1) The method comprises the following steps of collecting data information of coking or non-coking in a boiler, specifically: finding out the time intervals of severe coking and non-coking: finding out the time point of an effective coke falling event (a large coke falling event caused by artificial coke interference, natural coke falling and the like) from a coke falling platform account record input by a worker in a thermal power plant, then pushing the time point for 30 minutes forward until 3 hours before the time point is regarded as a time period with serious coking, and pushing the time point for 30 minutes backward until 3 hours after the time point is regarded as a time period without coking;
2) Selecting 123 measuring point temperature data which are coked or not coked in the same time period in the data information of the step 1), specifically: wall temperature data of corresponding measuring points in the time period determined in the step 1 are found out from a corresponding boiler wall temperature measuring point database of the thermal power plant and are divided into two parts: wall temperature data during severe coking and wall temperature data without coking;
3) Constructing a coked or non-coked convolutional neural network model, which comprises the following steps: an input layer, a convolutional layer, a downsampling layer, a full-link layer, and an output layer, the convolutional layer being divided into a convolutional layer C1 and a convolutional layer C2, the downsampling layer being divided into a downsampling layer S1 and a downsampling layer S2,
the method for inputting the layer comprises the following steps:
a. respectively supplementing the temperature data of 123 measuring points which are coked or not coked in the step 2) into a data matrix of 12 multiplied by 12, and supplementing the insufficient matrix elements with 0 (the purpose of supplementing 0 is to facilitate convolution operation during subsequent image feature extraction, so that the algorithm universality is improved);
b. normalizing each matrix element in the data matrix in the step a to obtain coking or non-coking analog image characteristics,
the normalized calculation formula is:
wherein: i, j are the indexes of the rows and columns of the data matrix in step a, x i，j Is a matrix element, v, of the normalized i, j position i，j To normalize the matrix elements of the previous i, j positions,
to normalize the smallest matrix element in the first j columns,
is the maximum matrix element in the first j columns of the normalization;
the convolutional layer C1: 3 convolution kernels are provided, the number of the convolution kernels is 3 x 3, and the analog image in the step b meets the conditions of min-max standardization, and meets the conditions that the mean value mu =0 and the variance sigma 2 Carrying out convolution operation on convolution kernels distributed in a standard plus-minus mode of =1 and the analog image features in the step b to obtain a feature map of the convolution layer C1, wherein in the sliding process of the convolution kernels, edges of the convolution kernels are filled with 0, the operation does not change the size of the image, the example outputs 12 × 12 × 3 data through the operation, three color channels (color is also one of the features of the image) of a common RGB color map, and the data generated at this time can be understood as the feature map of one 3 feature channels;
downsampling layer S1: the method comprises the following steps of (1) carrying out downsampling on a feature map of a convolutional layer C1 to obtain a feature map of a downsampling layer S1, carrying out abstract processing on image features by the downsampling layer to prevent overfitting and increase the robustness of a network, wherein the downsampling layer is essentially a convolution operation but has a simple function, blurring an image, fuzzifying unimportant features and strengthening important features (in the embodiment, a high-temperature area of a boiler is strengthened), in the embodiment, the main implementation process is a maximum pooling (maxpool) process, a 2 x 2 matrix window is utilized, row-column traversal is carried out with the sliding step length of 2, the maximum value in the 2 x 2 window is extracted to form a new matrix (for example, as shown in FIG. 2), and the same steps are carried out on 3 channels in sequence to obtain a 6 x 3 matrix feature map;
and (3) a convolutional layer C2: 5 convolution kernels are provided, the number of the convolution kernels is 3 x 3, under the condition that the feature map of the downsampling layer S1 is distributed according to the standard positive distribution, the feature map of the downsampling layer S1 is subjected to convolution operation, and a matrix feature map with a convolutional layer C2 of 6 x 5 is obtained;
downsampling layer S2: downsampling the feature map of the convolutional layer C2 to obtain a feature map of a downsampled layer S2 of 3 multiplied by 5;
the full articulamentum divide into full articulamentum 1 and full articulamentum 2 (also can only have full articulamentum 1), specifically is:
the method comprises the steps that (1) in a full connection layer, the number of neurons in the full connection layer is equal to the size of an image matrix after multilayer convolution operation is carried out on an image, each column of the image is extracted, and then the image is spliced into a one-dimensional vector 3 multiplied by 5=45, namely an input layer of the full connection layer 1 comprises 45 neurons;
selecting 15 neurons in the full connection layer 2, wherein the weight initialization mode of the full connection layer 2 is similar to the initialization of a convolution kernel and is a standard normal distribution matrix of 45 (the number of the neurons in the full connection layer 1) multiplied by 15 (the number of the neurons in the hidden layer);
an output layer: adopting a softmax function to perform probability output, including all neurons connected with a full connection layer, adopting a relu activation function for the convolutional layer and the full connection layer, adopting a random gradient descent method (SGD) as a network optimization mode, setting the learning rate to be 0.01, giving 50 training samples to a sample batch each time, performing iterative training on all data for 200 times, training a convolutional neural network model through a large amount of sample data, continuously correcting errors, reversely transmitting and updating a weight of a convolutional kernel, finally obtaining a relatively ideal convolutional neural network model, and storing parameters such as a model structure, a weight first change and the like;
4) Randomly selecting 123 measuring point temperature data in the same time period from a boiler data source acquired in real time, inputting the data into the coking or non-coking convolution neural network model in the step 3) to obtain the image characteristics of the measuring point temperature data, and judging the image characteristics to be in accordance with the coking image characteristics in the convolution neural network model, wherein the image characteristics are coking; and if the image features which are consistent with the non-coking in the convolutional neural network model are not coking, the image features are non-coking. As shown in fig. 3 and 4, where solid black is the complement and the 0 value is present in the real data, the higher the color intensity indicates the higher the temperature at the boiler measuring point, and different data can form different picture data characteristics.
Claims (5)
1. The boiler coking early warning method based on the convolutional neural network is characterized by comprising the following steps of:
1) Acquiring data information of coking or non-coking in the boiler;
2) Selecting temperature data of a plurality of measuring points which are coked or not coked in the same time period from the data information in the step 1);
3) Constructing a coked or non-coked convolutional neural network model, which comprises the following steps: an input layer, a convolutional layer, a downsampling layer, a full-link layer, and an output layer,
the method for inputting the layer comprises the following steps:
a. respectively supplementing the temperature data of the plurality of measuring points which are coked or not coked in the step 2) into an a x a data matrix, and supplementing the insufficient matrix elements with 0;
b. normalizing each matrix element in the data matrix in the step a to obtain coking or non-coking analog image characteristics,
the normalized calculation formula is:
wherein: i, j are the indexes of the rows and columns of the data matrix in step a, x i,j Is the matrix element, v, of the i, j position after normalization i,j Element of matrix for normalizing previous i, j positionPixel, or is present in>
For normalizing the smallest matrix element in the preceding j columns, <>
Is the maximum matrix element in the first j columns of the normalization;
4) Randomly selecting a plurality of measuring point temperature data in the same time period from a boiler data source acquired in real time, inputting the coking or non-coking convolution neural network model in the step 3) to obtain the image characteristics of the measuring point temperature data, and judging the image characteristics to be in accordance with the coking image characteristics in the convolution neural network model, wherein the image characteristics are coking; and if the image characteristics which accord with the non-coking in the convolutional neural network model are the non-coking, the image characteristics are the non-coking.
2. The boiler coking early warning method based on the convolutional neural network as claimed in claim 1, wherein in step 3), the convolutional layer: c, performing convolution operation on the analog image characteristics in the step b to obtain a characteristic diagram of the convolution layer;
down-sampling layer: down-sampling the feature map of the convolutional layer to obtain the feature map of the down-sampled layer;
full connection layer: flattening the feature map of the down-sampling layer to obtain a one-dimensional vector;
an output layer: and (5) adopting a softmax function to output the probability.
3. The boiler coking early warning method based on the convolutional neural network as claimed in claim 2, wherein in the step 3), the convolutional layer is divided into a convolutional layer C1 and a convolutional layer C2, and the downsampling layer is divided into a downsampling layer S1 and a downsampling layer S2, and the specific steps are as follows:
convolutional layer C1: performing convolution operation on the analog image characteristics in the step b to obtain a characteristic diagram of the convolution layer C1;
downsampling layer S1: down-sampling the feature map of the convolutional layer C1 to obtain a feature map of a down-sampled layer S1;
and (3) a convolutional layer C2: performing convolution operation on the feature map of the downsampling layer S1 to obtain a feature map of a convolution layer C2;
down-sampling layer S2: the feature map of the convolutional layer C2 is downsampled to obtain the feature map of the downsampled layer S2.
4. The boiler coking early warning method based on the convolutional neural network as claimed in claim 3, wherein in step 2), 123 measured point temperature data in the same time period are taken, and in the step a of inputting the measured point temperature data into the layer, the size of the data matrix is 12 x 12; convolution kernels of convolution layers C1 and C2 have a mean μ =0 and a variance σ 2 Normalized normal distribution of = 1; the convolution layer C1 has 3 convolution kernels, and the convolution kernels are 3 multiplied by 3; there are 5 convolution kernels in convolution layer C2, and the convolution kernel is 3 × 3.
5. The convolutional neural network-based boiler coking early warning method as claimed in claim 4, wherein the convolutional layer C1, the convolutional layer C2 and the full connection layer all adopt relu activation functions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910232761.2A CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910232761.2A CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109934417A CN109934417A (en) | 2019-06-25 |
| CN109934417B true CN109934417B (en) | 2023-04-07 |
Family
ID=66988260
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910232761.2A Active CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109934417B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113344909B (en) * | 2021-07-01 | 2023-12-08 | 中国石油大学(北京) | Method and device for identifying and displaying flame penetration height Wen Lvjing coking of thermal power boiler |
| CN118212408A (en) * | 2024-03-22 | 2024-06-18 | 国家能源集团科学技术研究院有限公司 | Hearth coking state judging method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105654067A (en) * | 2016-02-02 | 2016-06-08 | 北京格灵深瞳信息技术有限公司 | Vehicle detection method and device |
| CN107092906A (en) * | 2017-05-01 | 2017-08-25 | 刘至键 | A kind of Chinese traditional medicinal materials recognition device based on deep learning |
| CN109084613A (en) * | 2018-09-12 | 2018-12-25 | 东北电力大学 | Air cooling tubes condenser dust stratification status monitoring and cleaning control system and its regulation method based on convolutional neural networks and image recognition |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080116051A1 (en) * | 2006-09-29 | 2008-05-22 | Fisher-Rosemount Systems, Inc. | Main column bottoms coking detection in a fluid catalytic cracker for use in abnormal situation prevention |
| CN1975382A (en) * | 2006-12-07 | 2007-06-06 | 华南理工大学 | Coal slagging scorification trend fast monitoring instrument |
| CN102607009A (en) * | 2012-02-20 | 2012-07-25 | 华北电力大学 | Fouling monitoring system for convection heating surface of boiler |
| CN102928056B (en) * | 2012-11-22 | 2016-01-06 | 中国人民解放军国防科学技术大学 | The measuring method of hydrocarbon fuel coking amount |
| CN103064289B (en) * | 2012-12-19 | 2015-03-11 | 华南理工大学 | Multiple-target operation optimizing and coordinating control method and device of garbage power generator |
| CN104361153B (en) * | 2014-10-27 | 2017-11-03 | 中国石油大学(北京) | A kind of method for predicting RFCC settler coking amount |
| US9850434B2 (en) * | 2014-12-18 | 2017-12-26 | Exxonmobil Research And Engineering Company | Reduction of coking in FCCU feed zone |
| GB201512283D0 (en) * | 2015-07-14 | 2015-08-19 | Apical Ltd | Track behaviour events |
| CN105423273A (en) * | 2015-12-15 | 2016-03-23 | 天津鹰麟节能科技发展有限公司 | Spectroscopic boiler anti-coking system and control method |
| CN107256396A (en) * | 2017-06-12 | 2017-10-17 | 电子科技大学 | Ship target ISAR characteristics of image learning methods based on convolutional neural networks |
| CN107491781A (en) * | 2017-07-21 | 2017-12-19 | 国家电网公司 | A kind of crusing robot visible ray and infrared sensor data fusion method |
| CN107798336A (en) * | 2017-09-18 | 2018-03-13 | 广东电网有限责任公司东莞供电局 | Infrared temperature measurement image component identification method |
| CN108921893B (en) * | 2018-04-24 | 2022-03-25 | 华南理工大学 | Image cloud computing method and system based on online deep learning SLAM |
| CN108489912B (en) * | 2018-05-11 | 2019-08-27 | 东北大学 | A kind of coal constituent analysis method based on coal spectroscopic data |
-
2019
- 2019-03-26 CN CN201910232761.2A patent/CN109934417B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105654067A (en) * | 2016-02-02 | 2016-06-08 | 北京格灵深瞳信息技术有限公司 | Vehicle detection method and device |
| CN107092906A (en) * | 2017-05-01 | 2017-08-25 | 刘至键 | A kind of Chinese traditional medicinal materials recognition device based on deep learning |
| CN109084613A (en) * | 2018-09-12 | 2018-12-25 | 东北电力大学 | Air cooling tubes condenser dust stratification status monitoring and cleaning control system and its regulation method based on convolutional neural networks and image recognition |
Non-Patent Citations (2)
| Title |
|---|
| Ziwang Liu.Infrared Image Combined with CNN Based Fault Diagnosis for Rotating Machinery.IEEE Xplore.137-142. * |
| 孙自强,顾幸生,俞金寿,党晓恒.连续催化重整反应器结焦含量软测量.华东理工大学学报.(第05期),568-571. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109934417A (en) | 2019-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109523013B (en) | Air particulate matter pollution degree estimation method based on shallow convolutional neural network | |
| CN109615629B (en) | Convolutional neural network-based photovoltaic panel dust accumulation state image recognition system and analysis, regulation and control method thereof | |
| CN106548270B (en) | Photovoltaic power station power abnormity data identification method and device | |
| CN109934417B (en) | Boiler coking early warning method based on convolutional neural network | |
| CN112270449B (en) | Industrial system time delay determination and controlled quantity prediction method based on time correlation | |
| CN107749058A (en) | A kind of machine vision detection method and system of boiler tubing surface defect | |
| CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
| CN115271253A (en) | Water-wind power generation power prediction model construction method and device and storage medium | |
| CN114676622A (en) | Short-term photovoltaic power prediction method based on self-encoder deep learning model | |
| Fu et al. | Research on gas outburst prediction model based on multiple strategy fusion improved snake optimization algorithm with temporal convolutional network | |
| CN117498321A (en) | Distributed photovoltaic output prediction method, system and storage medium | |
| CN117010263A (en) | Residual life prediction method based on convolutional neural network and long-term and short-term memory network | |
| CN115163424A (en) | Wind turbine generator gearbox oil temperature fault detection method and system based on neural network | |
| CN113053536B (en) | Infectious disease prediction method, system and medium based on hidden Markov model | |
| Wang et al. | Combined ultra-short-term prediction method of PV power considering ground-based cloud images and chaotic characteristics | |
| CN105205560B (en) | Photovoltaic power supply power prediction method based on positive and negative error variable weights | |
| CN112884036A (en) | Boiler heating surface abnormal image identification method, marking method and system | |
| CN117849902A (en) | Method for forecasting and releasing precipitation through space-time double-correction high-resolution sub-seasonal grid set | |
| CN117096871A (en) | Wind power probability density prediction method based on space-time distribution | |
| CN114066819B (en) | Environmental corrosion severity identification method based on convolutional neural network deep learning | |
| CN113642784B (en) | Wind power ultra-short-term prediction method considering fan state | |
| CN113034465B (en) | Power equipment thermal fault monitoring method, device and medium based on infrared image | |
| CN114925931A (en) | Platform area load prediction method and system | |
| CN113836819A (en) | Bed temperature prediction method based on time sequence attention | |
| CN114492911A (en) | Short-term wind speed prediction method integrating data denoising and sparrow search algorithm and multi-layer extreme learning machine |
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 |