CN113344909B - Method and device for identifying and displaying flame penetration height Wen Lvjing coking of thermal power boiler - Google Patents

Abstract

The application discloses a method and a device for identifying and displaying the high flame penetration Wen Lvjing coking of a thermal power boiler, wherein the method comprises the steps of preprocessing an acquired high flame penetration Wen Lvjing coking image of the thermal power boiler; performing coking recognition based on a neural network prediction algorithm; quantitatively analyzing the identified coking, and calculating the coking fraction; early warning is carried out when the coking score is larger than a first preset threshold value, and intervention measures are adopted when the coking score is larger than a second preset threshold value; showing the coking condition. The method can realize quantitative analysis of the high-flame-permeability Wen Lvjing coking of the thermal power boiler and classification and display of coking by utilizing coking image recognition based on deep learning, and improves detection precision and efficiency. The application discloses a recognition display device for the high Wen Lvjing coking of the flame passing height of a thermal power boiler, which has the same advantages as the method.

Description

A method and device for identifying and displaying coking of flame-transparent high-temperature filters of thermal power boilers

Technical field

The invention belongs to the technical field of boiler coking monitoring in thermal power plants, and in particular relates to a method and device for identifying and displaying coking on a flame-transparent high-temperature filter of a thermal power boiler.

Background technique

Boiler coking is a common problem in coal-fired thermal power plants. It will destroy normal combustion conditions, reduce boiler efficiency, destroy normal water circulation, and even cause pipe explosion accidents. In severe cases, it will block the furnace outlet and force the furnace to shut down, affecting the Economic efficiency and safety of boiler operation. At present, when decoking the boiler furnace of a thermal power plant, it is necessary to send professional technicians to observe and predict the coking status through the fire viewing port, and then rely on experience to carry out decoking. Because the coking situation cannot be accurately judged under hot conditions, there is no suitable method Solution: It is impossible to make an accurate quantitative judgment on the location and degree of coking.

There is now a "boiler coking early warning method based on convolutional neural network", which collects temperature data of multiple measuring points in the same time period, obtains the image characteristics of the temperature data of the measuring points through the neural network, and determines whether it is coking, but it cannot determine whether it is coking. Coking type and coking state cannot be directly observed. This indirect measurement method belongs to the prediction of small sample data. The prediction of coking state and degree is not accurate enough and cannot be displayed intuitively.

Contents of the invention

In order to solve the above problems, the present invention provides a method and device for identifying and displaying coking in a flame-transmissive high-temperature filter of a thermal power boiler, which can realize quantitative analysis of coking and utilize coking image recognition based on deep learning to achieve classification and display of coking, improving Detection accuracy and efficiency.

A method for identifying and displaying coking of a flame-transparent high-temperature filter of a thermal power boiler provided by the invention includes:

Preprocess the collected images of thermal power boiler through flame high temperature filter focusing;

Coking identification based on neural network prediction algorithm;

Perform quantitative analysis on the identified coking and calculate the coking score;

When the coking score is greater than the first preset threshold, an early warning is issued, and when the coking score is greater than the second preset threshold, intervention measures are taken;

Shows the coking condition.

Preferably, in the above-mentioned method for identifying and displaying coking on flame-transparent high-temperature filters of thermal power boilers, the preprocessing of the collected coking images includes:

Reduce the amount of original image data based on the weighted average method;

Suppress image noise caused by the influence of the furnace environment based on the median filter method;

Outline the focused target object based on wavelet transform;

The original image is divided into focus and background objects based on an iterative method;

The focal image is first corroded and then expanded to eliminate small particle noise and smooth the focal boundary.

Preferably, in the above-mentioned method for identifying and displaying coking through flame high temperature filters of thermal power boilers, the coking identification based on the neural network prediction algorithm includes:

Create a coking image database with labels, the labels include coke block, high temperature coking, other and normal;

Split the focal image database, taking 60% of the training set and 40% of the test set;

Create an enhanced focus image database by setting the image size and then flipping it, load it into the Alex network, and modify the number of fully connected layer output categories;

Set the learning rate and loop parameters and start training;

Read the focus images in the test set, perform prediction and classification, and obtain accuracy information;

The neural network is optimized by optimizing the learning rate and enhancing the distribution of the data set.

Preferably, in the above-mentioned method for identifying and displaying coking in a flame-transmitting high-temperature filter of a thermal power boiler, the quantitative analysis of the identified coking includes:

Select a calibration object whose actual size is known to be d;

Measure the number of pixels D1 and D2 of the calibration object and the coked object in the image, and determine the calibration coefficient d/D1;

Quantitatively calculate the focal area in the target image.

Preferably, in the above-mentioned method for identifying and displaying coking in a flame-transmitting high-temperature filter of a thermal power boiler, the display of the coking situation includes:

Displays coking images, coking classification results, coking quantitative analysis results, and coking treatment suggestions.

A thermal power boiler flame-transmitting high-temperature filter coking identification and display device provided by the invention includes:

The preprocessing component is used to preprocess the collected images of the thermal power boiler through flame high temperature filter focusing;

Identification component, used for coking identification based on neural network prediction algorithm;

The analysis and calculation component is used to quantitatively analyze the identified coking and calculate the coking score;

An early warning component, configured to provide an early warning when the coking score is greater than a first preset threshold, and to take intervention measures when it is greater than a second preset threshold;

A display component is used to display the coking situation.

Preferably, in the above-mentioned flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the pretreatment component includes:

Data volume reduction unit, used to reduce the volume of original image data based on the weighted average method;

Noise suppression unit, used to suppress image noise caused by environmental effects in the furnace based on the median filter method;

An outlining unit is used to outline the focused target object based on wavelet transform;

A background object distinguishing unit, used to divide the original image into focus and background objects based on an iterative method;

The boundary smoothing unit is used to first erode and then expand the focus image to eliminate small particle noise and smooth the focus boundary.

Preferably, in the above-mentioned flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the identification component includes:

A database creation unit configured to create a labeled coking image database, where the labels include coke block, high temperature coking, other and normal;

A segmentation unit used to segment the focal image database, taking 60% of the training set and 40% of the test set;

The enhanced database creation unit is used to create an enhanced focused image database by setting the image size and then flipping it, and loads the Alex network to modify the number of fully connected layer output categories;

The training unit is used to set the learning rate and loop parameters and start training;

A prediction classification unit, used to read the focused images in the test set, perform prediction classification, and obtain accuracy information;

A neural network optimization unit is used to optimize the neural network by optimizing the learning rate and enhancing the distribution of the data set.

Preferably, in the above-mentioned thermal power boiler flame-transmitting high-temperature filter coking identification and display device, the analysis and calculation component includes:

The calibration object selection unit is used to select the calibration object whose actual size is known to be d;

The pixel number measuring unit is used to measure the pixel numbers D1 and D2 of the calibration object and the coking object in the image, and determine the calibration coefficient d/D1;

The coking area calculation unit is used to quantitatively calculate the coking area in the target image.

Preferably, in the above-mentioned coking identification and display device of a flame-transmitting high-temperature filter for thermal power boilers, the display component is specifically used to display coking images, coking classification results, coking quantitative analysis results and coking treatment suggestions.

It can be seen from the above description that the above-mentioned flame-transmitting high-temperature filter coking identification and display method for thermal power boilers provided by the present invention includes preprocessing the collected flame-transmitting high-temperature filter coking images of thermal power boilers; performing coking identification based on a neural network prediction algorithm; The identified coking is quantitatively analyzed and the coking score is calculated; when the coking score is greater than the first preset threshold, an early warning is issued, and when it is greater than the second preset threshold, intervention measures are taken; the coking situation is displayed, so that it can be realized Quantitative analysis of coking and the use of coking image recognition based on deep learning to classify and display coking, improving detection accuracy and efficiency. The above-mentioned flame-transmitting high-temperature filter coking identification and display device for thermal power boilers provided by the present invention has the same advantages as the above-mentioned method.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only These are embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without exerting creative efforts.

Figure 1 is a schematic diagram of an embodiment of a method and device for identifying and displaying coking of a flame-transmitting high-temperature filter of a thermal power boiler provided by the present invention;

Figure 2 is a schematic diagram of the visualization software module used in the coking identification and display method of the flame-transmitting high-temperature filter of a thermal power boiler;

Figure 3 is a schematic diagram of an embodiment of a flame-transmitting high-temperature filter coking identification and display device for a thermal power boiler provided by the present invention.

Detailed ways

The core of the present invention is to provide a method and device for identifying and displaying coking of a flame-transmitting high-temperature filter of a thermal power boiler, which can realize quantitative analysis of coking of a flame-transmitting high-temperature filter of a thermal power boiler and utilize coking image recognition based on deep learning to achieve coking classification. and display to improve detection accuracy and efficiency.

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

An embodiment of a method for identifying and displaying coking of a flame-transparent high-temperature filter of a thermal power boiler provided by the invention is shown in Figure 1. Figure 1 is a schematic diagram of an embodiment of a method of identifying and displaying coking of a flame-transparent high-temperature filter of a thermal power boiler provided by the invention. , the method includes the following steps:

S1: Preprocess the collected image of thermal power boiler through flame high temperature filter focusing;

It should be noted that after preprocessing, the coked part of the flame-transmitting high-temperature filter of the thermal power boiler can be made more prominent and easier to identify, thereby improving the identification efficiency.

S2: Coking identification based on neural network prediction algorithm;

Specifically, this neural network prediction algorithm can be trained based on the Alex network in MATLAB.

S3: Perform quantitative analysis on the identified coking and calculate the coking score;

Specifically, the coking status can be scored according to the coking type and coking area. For example, it can be divided into red, orange, yellow, and green levels, where red indicates danger and green indicates safety.

S4: Provide an early warning when the coking score is greater than the first preset threshold, and take intervention measures when it is greater than the second preset threshold;

Specifically, it can be judged whether to issue an early warning based on the analyzed coking score, which facilitates the operator to respond promptly to the coking situation and avoid making the situation more serious if the time is too long. It can be, but is not limited to, set to a value higher than 70. If the score exceeds 90, manual intervention will be required, and if the score exceeds 90, intervention measures will be taken to avoid severe situations such as excessive coking.

S5: Displays the coking situation.

Specifically, but not limited to, the host computer screen can be used to display the coking situation. Various coking parameters, images, analysis results, etc. can be displayed at the same time, allowing the operator to better grasp the real-time situation of coking.

As can be seen from the above description, the embodiment of the method for identifying and displaying the coking of the flame-transmitting high-temperature filter of a thermal power boiler provided by the present invention includes preprocessing the collected image of the coking of the flame-transmitting high-temperature filter of the thermal power boiler; based on a neural network prediction algorithm. Coking identification; quantitatively analyze the identified coking and calculate the coking score; issue an early warning when the coking score is greater than the first preset threshold, and take intervention measures when it is greater than the second preset threshold; display the coking situation, so it can be achieved Quantitative analysis of coking and the use of coking image recognition based on deep learning to classify and display coking, improving detection accuracy and efficiency.

In a specific embodiment of the above-mentioned flame-transmitting high-temperature filter coking identification and display method for thermal power boilers, preprocessing the collected flame-transmitting high-temperature filter coking images of thermal power boilers may include:

Reduce the amount of original image data based on the weighted average method, which is a grayscale process, which can speed up processing;

Based on the median filtering method, the image noise caused by the influence of the furnace environment is suppressed, which is a smooth denoising process, which can improve the image quality;

Based on wavelet transform, the focused target object is outlined to make its features more obvious, which is an edge detection process;

The original image is divided into focus and background objects based on the iterative method, which is a binarization process;

The focal image is first corroded and then expanded to eliminate small particle noise and smooth the focal boundary. This is an open operation process.

In another specific embodiment of the above-mentioned method for identifying and displaying coking in a flame-transparent high-temperature filter of a thermal power boiler, identifying coking based on a neural network prediction algorithm may include the following steps:

Create a labeled coking image database, with labels including coke block, high temperature coking, other and normal;

Split the focused image database, taking 60% of the training set and 40% of the test set;

Create an enhanced focus image database by setting the image size and then flipping it, load it into the Alex network, and modify the number of fully connected layer output categories;

Set the learning rate and loop parameters, start training, and save the trained network;

Read the focus images in the test set, perform prediction and classification, and obtain accuracy information;

Optimize the neural network by optimizing the learning rate and enhancing the distribution of the data set to continuously improve accuracy.

In another specific embodiment of the above-mentioned method for identifying and displaying coking in a flame-transmitting high-temperature filter of a thermal power boiler, quantitative analysis of the identified coking may include:

Select a calibration object whose actual size is known to be d;

Measure the number of pixels D1 and D2 of the calibration object and the coked object in the image, and determine the calibration coefficient d/D1;

Quantitatively calculate the focal area in the target image.

In other words, in this way, the specific size of coking can be identified. When it reaches a certain threshold, an early warning will be issued and timely processing can be done to avoid greater risks caused by coking.

In a preferred embodiment of the above-mentioned method for identifying and displaying coking in a flame-transparent high-temperature filter of a thermal power boiler, displaying coking conditions may include:

Displays coking images, coking classification results, coking quantitative analysis results, and coking treatment suggestions.

Specifically, two interfaces can be displayed in this step: one is the image processing interface, which includes various image processing algorithms to be selected, calculation of the focal block area, display of the original image and the processed image, etc.; the second is the image processing interface. The fault diagnosis interface includes the display of coking images, the display of coking image classification results, the display of diagnosis results, the display of scores, the display of alarm indicators and the display of suggested treatment methods, etc.

Referring to Figure 2, Figure 2 is a schematic diagram of the visual software module used in the coking identification and display method of the thermal power boiler through flame high temperature filter, which includes a coking image reading module, a coking image processing module, a coking image classification and identification module, and a coking image display module. Coking image quantitative analysis module, alarm module, result display module and suggestion prompt module. After the coking image is analyzed and processed by the coking image reading module and the image processing module, the coking image display module displays the original image and the processed image respectively. Real-time and intuitive display is performed. At the same time, the coking image is handed over to the coking image classification and identification module and the coking quantitative analysis module for classification, identification and quantitative analysis. The results are displayed through the classification result display module and the quantitative analysis result display module. The alarm module is based on the classification. The results and quantitative analysis results are subjected to threshold judgment and scoring. If the score exceeds the set threshold, the red light will light up and an alarm will be issued. At this time, a new picture window will be opened, and the focused single frame image at this moment will be enlarged and displayed on the home page and automatically saved. Go to the coking image folder. If the score does not exceed the threshold, the alarm module will not make subsequent responses. The suggestion prompt module will also make corresponding operation suggestions based on the results of coking classification identification and the results of quantitative analysis.

In summary, the above-mentioned coking identification and display method of the flame-transmitting high-temperature filter of thermal power boilers can carry out coking identification classification, quantitative analysis, provide a basis for formulating decoking strategies, overcome human factors, and ensure the safe and economical operation of the boiler.

An embodiment of a coking identification and display device for a flame-transparent high-temperature filter of a thermal power boiler provided by the invention is shown in Figure 3. Figure 3 is a schematic diagram of an embodiment of a coking identification and display device for a flame-transparent high-temperature filter of a thermal power boiler provided by the invention. , can be integrated into the host computer, including:

The preprocessing component 301 is used to preprocess the collected coked images of the thermal power boiler through the flame high temperature filter. It should be noted that after preprocessing, the coked parts can be made more prominent and easy to identify, thereby improving the identification efficiency;

The identification component 302 is used for coking identification based on a neural network prediction algorithm. Specifically, this neural network prediction algorithm can be trained based on the Alex network in MATLAB;

The analysis and calculation component 303 is used to quantitatively analyze the identified coking and calculate the coking score. Specifically, the coking status can be scored according to the type of coking and the coking area. For example, it can be divided into red, orange, yellow, and green. Level, where red indicates danger and green indicates safety;

The early warning component 304 is used to issue an early warning when the coking score is greater than the first preset threshold, and to take intervention measures when it is greater than the second preset threshold. Specifically, it can be judged whether to issue an early warning based on the analyzed coking score, which is convenient. The operator responds to the coking situation in a timely manner to avoid the situation becoming more serious if the time is too long. It can be, but is not limited to, set to manual intervention if the score exceeds 70, and intervention measures will be taken if the score exceeds 90 to avoid coking. Too large and other adverse conditions;

The display component 305 is used to display the coking situation. Specifically, it can be, but is not limited to, using a host computer screen to display the coking situation. Various coking parameters, images, analysis results, etc. can be displayed at the same time to allow the operator to better understand the coking situation. Understand the real-time situation of coking.

In a specific embodiment of the above-described flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the preprocessing component may include:

The data volume reduction unit is used to reduce the volume of original image data based on the weighted average method, which is also a grayscale process, which can speed up the processing;

The noise suppression unit is used to suppress the image noise caused by the environmental influence in the furnace based on the median filter method. This is also a smooth denoising process, which can improve the image quality.

The outlining unit is used to outline the focused target object based on wavelet transform, which is also an edge detection process;

The background object distinction unit is used to divide the original image into focus and background objects based on an iterative method, which is a binarization process;

The boundary smoothing unit is used to first erode and then expand the focal image to eliminate small particle noise and smooth the focal boundary. This is an open operation process.

In another specific embodiment of the above-described flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the identification components may include:

A database creation unit used to create a labeled coking image database, where the labels include coke block, high temperature coking, other and normal;

The segmentation unit is used to segment the focal image database, taking 60% of the training set and 40% of the test set;

The enhanced database creation unit is used to create an enhanced focused image database by setting the image size and then flipping it, and loads the Alex network to modify the number of fully connected layer output categories;

The training unit is used to set the learning rate and loop parameters and start training;

The prediction classification unit is used to read the focus images in the test set, perform prediction classification, and obtain accuracy information;

The neural network optimization unit is used to optimize the neural network by optimizing the learning rate and enhancing the distribution of the data set.

In another specific embodiment of the above-described flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the analysis and calculation components may include:

The calibration object selection unit is used to select the calibration object whose actual size is known to be d;

The pixel number measuring unit is used to measure the pixel numbers D1 and D2 of the calibration object and the coking object in the image, and determine the calibration coefficient d/D1;

The coking area calculation unit is used to quantitatively calculate the coking area in the target image.

In other words, in this way, the specific size of coking can be identified. When it reaches a certain threshold, an early warning will be issued and timely processing can be done to avoid greater risks caused by coking.

In a preferred embodiment of the above-mentioned flame-transmitting high-temperature filter coking identification and display device for thermal power boilers, the display component can be used to display coking images, coking classification results, coking quantitative analysis results, and coking treatment suggestions.

Two interfaces can be displayed: one is the image processing interface, which includes various image processing algorithms to be selected, calculation of the focal block area, display of original images and processed images, etc.; the other is the fault diagnosis interface, including identification Display of coking images, display of coking image classification results, display of diagnosis results, display of scores, display of alarm lights and display of suggested treatment methods, etc.

The above description of the disclosed embodiments enables those skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be practiced in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (6)

1. A method for identifying and displaying coking of a flame-transparent high-temperature filter of a thermal power boiler, which is characterized by including: Preprocess the collected images of thermal power boiler through flame high temperature filter focusing; Coking identification based on neural network prediction algorithm; Perform quantitative analysis on the identified coking and calculate the coking score; When the coking score is greater than the first preset threshold, an early warning is issued, and when the coking score is greater than the second preset threshold, intervention measures are taken; Display the coking condition; The preprocessing of the collected image of the thermal power boiler through the flame high-temperature filter focusing includes: Reduce the amount of original image data based on the weighted average method; Suppress image noise caused by the influence of the furnace environment based on the median filter method; Outline the focused target object based on wavelet transform; The original image is divided into focus and background objects based on an iterative method; The focal image is first corroded and then expanded to eliminate small particle noise and smooth the focal boundary; The coking identification based on the neural network prediction algorithm includes: Create a coking image database with labels, the labels include coke block, high temperature coking, other and normal; Split the focal image database, taking 60% of the training set and 40% of the test set; Create an enhanced focus image database by setting the image size and then flipping it, load it into the Alex network, and modify the number of fully connected layer output categories; Set the learning rate and loop parameters and start training; Read the focus images in the test set, perform prediction and classification, and obtain accuracy information; The neural network is optimized by optimizing the learning rate and enhancing the distribution of the data set. 2. The method for identifying and displaying coking in flame-transmitting high-temperature filters of thermal power boilers according to claim 1, characterized in that the quantitative analysis of the identified coking includes: Select a calibration object whose actual size is known to be d; Measure the number of pixels D1 and D2 of the calibration object and the coked object in the image, and determine the calibration coefficient d/D1; Quantitatively calculate the focal area in the target image. 3. The method for identifying and displaying coking of a flame-transmitting high-temperature filter of a thermal power boiler according to claim 2, characterized in that the display of the coking situation includes: Displays coking images, coking classification results, coking quantitative analysis results, and coking treatment suggestions. 4. A thermal power boiler flame-transmitting high-temperature filter coking identification and display device, which is characterized in that it includes: The preprocessing component is used to preprocess the collected images of the thermal power boiler through flame high temperature filter focusing; Identification component, used for coking identification based on neural network prediction algorithm; The analysis and calculation component is used to quantitatively analyze the identified coking and calculate the coking score; An early warning component, configured to provide an early warning when the coking score is greater than a first preset threshold, and to take intervention measures when it is greater than a second preset threshold; Display component, used to display the coking situation; The preprocessing components include: Data volume reduction unit, used to reduce the volume of original image data based on the weighted average method; Noise suppression unit, used to suppress image noise caused by environmental effects in the furnace based on the median filter method; An outlining unit is used to outline the focused target object based on wavelet transform; A background object distinguishing unit, used to divide the original image into focus and background objects based on an iterative method; The boundary smoothing unit is used to first erode and then expand the focus image to eliminate small particle noise and smooth the focus boundary; The identification components include: A database creation unit configured to create a labeled coking image database, where the labels include coke block, high temperature coking, other and normal; A segmentation unit used to segment the focal image database, taking 60% of the training set and 40% of the test set; The enhanced database creation unit is used to create an enhanced focus image database by setting the image size and then flipping it, and loads the Alex network to modify the number of fully connected layer output categories; The training unit is used to set the learning rate and loop parameters and start training; A prediction classification unit, used to read the focused images in the test set, perform prediction classification, and obtain accuracy information; A neural network optimization unit is used to optimize the neural network by optimizing the learning rate and enhancing the distribution of the data set. 5. The thermal power boiler flame-transmitting high-temperature filter coking identification and display device according to claim 4, characterized in that the analysis and calculation component includes: The calibration object selection unit is used to select the calibration object whose actual size is known to be d; The pixel number measuring unit is used to measure the pixel numbers D1 and D2 of the calibration object and the coking object in the image, and determine the calibration coefficient d/D1; The coking area calculation unit is used to quantitatively calculate the coking area in the target image. 6. The thermal power boiler through-flame high-temperature filter coking identification display device according to claim 5, characterized in that the display component is specifically used to display coking images, coking classification results, coking quantitative analysis results and coking treatment suggestions.