Boiler soot blowing control method and related device

Abstract

The invention provides a boiler soot blowing control method and a related device, and relates to the technical field of boiler operation control. Performing image processing on the initial image of the boiler heating surface to obtain a processed image, performing mapping operation on pixel coordinates of the processed image and actual reference coordinates of the boiler heating surface to obtain a mapping result, and obtaining a region image corresponding to the soot blower based on the mapping result, carrying out dust deposition identification operation on the regional image to obtain a dust deposition identification result, calculating regional attribute information of the regional image according to the dust deposition identification result, determining a fuzzy membership function boundary value corresponding to the regional attribute information, constructing a self-adaptive fuzzy membership function by using the fuzzy membership function boundary value, and obtaining a self-adaptive fuzzy membership function; and determining a fuzzy quantity corresponding to the regional attribute information by using an adaptive fuzzy membership function, determining a soot blowing parameter of a soot blower corresponding to the regional image according to the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, and controlling the corresponding soot blower to perform soot blowing operation according to the soot blowing parameter.

Description

Technical Field

[0001] This application relates to the field of boiler operation control technology, and in particular to a boiler soot blowing control method and related devices. Background Technology

[0002] As the core equipment of thermal power generation, the thermal efficiency of boilers directly affects power generation efficiency and operating costs. Soot blowing is a key part of daily boiler maintenance. By removing ash buildup on the boiler's heating surfaces, heat exchange efficiency can be effectively improved and fuel consumption reduced. However, excessive or improper soot blowing can lead to wear on the heating surfaces, increasing maintenance costs.

[0003] Therefore, how to achieve precise control of boiler soot blowing is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of the above problems, this application provides a boiler soot blowing control method and related device to achieve precise control of boiler soot blowing. The specific solution is as follows:

[0005] The first aspect of this application provides a boiler soot blowing control method, comprising:

[0006] An image acquisition operation is performed on the boiler heating surface to obtain an initial image;

[0007] The initial image is processed to obtain a processed image. The pixel coordinates in the processed image are mapped to the actual reference coordinates of the boiler heating surface to obtain a mapping result. Based on the mapping result, the processed image is segmented according to the sootblowing area covered by each sootblower to obtain the region image corresponding to the sootblower. The image processing operation includes at least distortion correction, grayscale conversion, color enhancement, and noise reduction filtering.

[0008] A gray accumulation recognition operation is performed on the image of the region to obtain a gray accumulation recognition result; the gray accumulation recognition result includes gray accumulation regions and non-gray accumulation regions.

[0009] Based on the dust accumulation recognition result, calculate the regional attribute information of the region image; the regional attribute information includes the region dust accumulation coverage, the region average dust accumulation thickness, and the region roughness.

[0010] Determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function;

[0011] Based on the fuzziness amount corresponding to the regional attribute information and the fuzzy rule library, the blowing parameters of the soot blower corresponding to the regional image are determined, and the soot blower is controlled to perform the blowing operation according to the blowing parameters.

[0012] In one possible implementation, a gray-accumulation recognition operation is performed on the region image to obtain a gray-accumulation recognition result, including:

[0013] The image of the region is segmented into gray areas and non-gray areas to obtain the gray area recognition result.

[0014] In one possible implementation, based on the gray accumulation recognition result, the region attribute information of the region image is calculated, including:

[0015] Calculate the proportion of pixels located in the gray area to the total number of pixels in the region image to obtain the region gray coverage rate;

[0016] Calculate the average gray value of the region image to obtain the average gray thickness of the region.

[0017] Calculate the region texture uniformity, region grayscale difference, and region texture complexity of the region image;

[0018] The real-time roughness of the region image is calculated based on the region texture uniformity, the region grayscale difference, and the region texture complexity.

[0019] The regional dust accumulation roughness of the image is calculated based on the real-time roughness of the region and the reference roughness.

[0020] In one possible implementation, determining the fuzzy membership function boundary values ​​corresponding to the region attribute information includes:

[0021] Clustering operations are performed on the region attribute information to obtain initial boundary values;

[0022] Based on the historical soot blowing data of the soot blower, a correction value is obtained;

[0023] The sum of the initial boundary value and the correction value is used as the boundary value of the fuzzy membership function.

[0024] In one possible implementation, a correction value is obtained based on the historical soot blowing data of the soot blower, including:

[0025] Obtain the actual soot blowing effect feedback index of the soot blower during the previous soot blowing operation and the expected soot blowing effect feedback index corresponding to the actual soot blowing effect feedback index.

[0026] Obtain the minimum and maximum area attribute information of the sootblower during historical sootblowing operations;

[0027] Based on the difference between the actual soot blowing effect feedback index and the expected soot blowing effect feedback index, the difference between the regional attribute information and the minimum regional attribute information, and the difference between the maximum regional attribute information and the minimum regional attribute information, a correction value is calculated.

[0028] In one possible implementation, the fuzzy quantity corresponding to the regional attribute information is determined using the adaptive fuzzy membership function, including:

[0029] The region attribute information is fuzzified using the adaptive fuzzy membership function to obtain the fuzzy quantity corresponding to the region attribute information.

[0030] In one possible implementation, the blowing parameters of the soot blower corresponding to the region image are determined based on the fuzziness amount corresponding to the region attribute information and the fuzzy rule base, including:

[0031] Obtain a pre-established fuzzy rule base;

[0032] Based on the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, the rule activation degree is calculated, and fuzzy inference is performed based on the rule activation degree to obtain the output fuzzy quantity.

[0033] Using the output blur quantity, a deblurring operation is performed using the centroid method to obtain the blowing parameters of the soot blower corresponding to the region image; the blowing parameters include at least the blowing intensity coefficient and start / stop commands.

[0034] A second aspect of this application provides a boiler soot blowing control device, comprising:

[0035] The image acquisition module is used to acquire images of the boiler's heating surface to obtain initial images.

[0036] The image processing module is used to perform image processing operations on the initial image to obtain a processed image, map the pixel coordinates in the processed image to the actual reference coordinates of the boiler heating surface to obtain a mapping result, and based on the mapping result, perform image segmentation operations on the processed image according to the soot blowing area covered by each soot blower to obtain the region image corresponding to the soot blower; the image processing operations include at least distortion correction, grayscale conversion, color enhancement, and noise reduction filtering;

[0037] The dust accumulation recognition module is used to perform dust accumulation recognition on the region image to obtain a dust accumulation recognition result; the dust accumulation recognition result includes a dust accumulation region and a non-dust accumulation region.

[0038] The information calculation module is used to calculate the regional attribute information of the region image based on the dust accumulation recognition result; the regional attribute information includes the region dust accumulation coverage, the region average dust accumulation thickness, and the region roughness.

[0039] The fuzzy quantity determination module is used to determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function.

[0040] The soot blowing control module is used to determine the soot blowing parameters of the soot blower corresponding to the region image based on the fuzziness amount corresponding to the region attribute information and the fuzzy rule library, and control the corresponding soot blower to perform soot blowing operation according to the soot blowing parameters.

[0041] A third aspect of this application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0042] The memory is used to store computer programs;

[0043] The processor is used to execute the computer program so that the electronic device can implement the above-described boiler soot blowing control method.

[0044] A fourth aspect of this application provides a computer storage medium carrying one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the above-described boiler soot blowing control method.

[0045] By means of the above technical solution, this application provides a boiler soot blowing control method and related device, which involves image acquisition of the boiler heating surface to obtain an initial image, image processing such as distortion correction, grayscale conversion, color enhancement, and noise reduction filtering on the initial image to obtain a processed image, mapping the pixel coordinates of the processed image to the actual reference coordinates of the boiler heating surface to obtain a mapping result, segmenting the processed image according to the soot blowing area covered by each soot blower to obtain the area image corresponding to the soot blower, and performing ash accumulation recognition on the area image to obtain the ash accumulation recognition result; the ash accumulation... The gray area recognition result includes gray-accumulated areas and non-gray-accumulated areas. Based on the gray area recognition result, the regional attribute information of the regional image is calculated. The regional attribute information includes gray coverage, average gray thickness, and regional roughness. The boundary values ​​of the fuzzy membership function corresponding to the regional attribute information are determined. An adaptive fuzzy membership function is constructed using the boundary values ​​of the fuzzy membership function. The fuzzy quantity corresponding to the regional attribute information is determined using the adaptive fuzzy membership function. Based on the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, the sootblowing parameters of the sootblower corresponding to the regional image are determined. The sootblower is controlled to perform sootblowing operation according to the sootblowing parameters. In this application, the gray accumulation situation is analyzed based on the real-time operating data of the boiler heating surface to obtain accurate sootblowing parameters, thereby enabling precise control of boiler sootblowing. Attached Figure Description

[0046] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0047] Figure 1 A flowchart of a boiler soot blowing control method provided in this application;

[0048] Figure 2 A flowchart of a method for calculating regional attribute information provided in this application;

[0049] Figure 3 A flowchart of a method for calculating the boundary values ​​of a fuzzy membership function is provided in this application;

[0050] Figure 4 A flowchart of a method for determining soot blowing parameters provided in this application;

[0051] Figure 5 A schematic diagram illustrating a scenario for a boiler soot blowing control method provided in this application;

[0052] Figure 6This application provides a schematic diagram of the structure of a boiler soot blowing control device;

[0053] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0054] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0055] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0056] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0057] As the core equipment in thermal power generation, the boiler's thermal efficiency directly affects power generation efficiency and operating costs. Soot blowing is a crucial part of routine boiler maintenance. By removing ash buildup on the boiler's heating surfaces, heat exchange efficiency can be effectively improved and fuel consumption reduced. However, excessive or improper soot blowing can lead to wear on the heating surfaces, increasing maintenance costs. Therefore, precise control of boiler soot blowing is essential for improving overall operational efficiency.

[0058] Currently, soot blowing systems based on timed and pressure differential control can be used for boiler soot blowing. This method uses preset time intervals and the pressure difference between the boiler inlet and outlet as conditions to trigger soot blowing. When the set time or pressure difference threshold is reached, the system automatically starts the soot blowing program. While this solution is simple and direct, it lacks real-time monitoring of the actual ash accumulation in the boiler, cannot accurately locate areas with severe ash accumulation, and cannot adjust the soot blowing strategy in real time according to the actual operating conditions and ash accumulation. This can easily lead to insufficient or excessive soot blowing, affecting boiler efficiency and lifespan. Furthermore, this soot blowing method not only increases energy consumption but may also accelerate the wear of the boiler's heating surfaces, reducing the overall service life of the boiler and wasting resources. In addition, this soot blowing method has poor adaptability; different boilers, different fuels, and different operating environments can all affect the speed and distribution of ash accumulation, and the system's lack of consideration for these factors significantly reduces its effectiveness in practical applications.

[0059] In addition, an ash accumulation monitoring system based on infrared thermal imaging technology can be used for soot blowing. Specifically, an infrared thermal imager scans the boiler's heating surface, and the degree of ash accumulation is indirectly determined by analyzing changes in the temperature distribution of the heating surface. When an abnormally high local temperature is detected, it is considered a severely ash-accumulated area, triggering a soot blowing operation. Although this method improves the targeting of soot blowing control, the ash accumulation monitoring system based on infrared thermal imaging technology is greatly affected by environmental factors. While infrared thermal imaging technology can reflect the temperature distribution of the heating surface, it is easily interfered with by factors such as smoke and dust inside the boiler and external light, leading to decreased measurement accuracy or even malfunction. Furthermore, infrared imaging technology relies solely on temperature information to determine the degree of ash accumulation, ignoring crucial information such as the physical morphology and distribution range of the ash, resulting in incomplete and inaccurate assessment results. Moreover, infrared thermal imaging technology is relatively complex, requiring professional technical support and maintenance, increasing the overall operating cost and technical difficulty of the system.

[0060] In summary, while the two aforementioned boiler soot blowing control methods have achieved automation to some extent, they still have significant shortcomings in terms of real-time performance, accuracy, adaptability, efficiency, and technical complexity. This application proposes a novel image recognition-based boiler soot blowing control method to overcome these deficiencies and achieve more efficient, accurate, and economical boiler soot blowing control.

[0061] Specifically, the boiler soot blowing control method directly captures images of the boiler's heating surface using a high-resolution, high-temperature resistant camera. It then employs advanced image processing and machine learning algorithms to identify the areas and extent of ash accumulation, enabling real-time and accurate assessment of the degree of ash accumulation. Based on this, it automatically adjusts the soot blowing strategy to optimize boiler operating efficiency and reduce maintenance costs, demonstrating high precision.

[0062] Based on the above, one embodiment of this application provides a boiler soot blowing control method that can capture images of ash accumulation on the boiler's heating surface in real time. By comprehensively utilizing image processing algorithms and intelligent control logic, it accurately identifies the ash accumulation area and degree, and dynamically adjusts the soot blowing strategy accordingly. This application overcomes the shortcomings of traditional soot blowing control methods, achieving precise, intelligent, and efficient boiler soot blowing operations, and providing strong support for the safe, stable, and efficient operation of the boiler.

[0063] Reference Figure 1 A boiler soot blowing control method may include:

[0064] S11. Perform image acquisition on the boiler heating surface to obtain the initial image.

[0065] In practical implementation, cameras can be used to acquire images of the boiler heating surfaces in thermal power plants. When selecting and deploying cameras, high-resolution, wide-angle, and high-temperature-resistant industrial-grade cameras should be chosen to ensure they can clearly capture images of the boiler heating surfaces under conditions of strong radiation and high levels of coal dust. Furthermore, high-definition cameras should have an air / water cooling design. Camera placement should be determined based on the actual boiler soot blowing requirements, such as the front and rear walls of the furnace, side walls, water-cooled walls, high-temperature heating surfaces, and economizer areas. The number of cameras should be determined based on the furnace type and size, with high-definition cameras staggered between adjacent layers. The deployment principle is to receive as much ash accumulation information as possible from the entire furnace, facilitating comprehensive control of the soot blowing effect. In actual scenarios, soot blowers are installed inside the boiler. The initial image captured by one camera includes the soot blowing area covered by at least one soot blower. The soot blowing area covered by one soot blower can be captured by multiple cameras. Subsequently, the area image of all the soot blowing areas covered by the soot blowers captured by the cameras can be randomly or according to certain rules, and this area image can be used as the corresponding area image for that soot blower.

[0066] The initial image output from the camera undergoes preprocessing circuits such as amplification and filtering to improve signal stability and anti-interference capabilities. Simultaneously, a high-speed A / D converter (Analog to Digital) is used to convert the analog signal into a digital signal for easier subsequent processing.

[0067] S12. Perform image processing on the initial image to obtain the processed image. Map the pixel coordinates in the processed image to the actual reference coordinates of the boiler heating surface to obtain the mapping result. Based on the mapping result, perform image segmentation on the processed image according to the soot blowing area covered by each soot blower to obtain the area image corresponding to the soot blower.

[0068] Image processing operations include at least distortion correction, grayscale conversion, color enhancement, and noise reduction filtering.

[0069] In practice, the images captured by the camera are transmitted to the backend image processing module for image processing (or image preprocessing). The image processing process is as follows:

[0070] Considering the potential distortion during shooting, it is necessary to calibrate the camera intrinsic parameters and distortion coefficients using OpenCV functions to perform distortion correction on the initial image. Furthermore, the initial image should undergo preprocessing such as grayscale conversion, image enhancement (color enhancement), and noise reduction filtering to ensure the geometric reliability and quality of the image itself.

[0071] Then, using OpenCV functions and calibration points (selecting multiple points with clear and stable geometric features), the mapping matrix H is solved. By mapping the pixel coordinates in the processed image to the actual reference coordinates of the actual position of the boiler heating surface, the mapping formula can be obtained, namely:

[0072] ;

[0073] Where, x t y t The actual reference coordinates are the horizontal and vertical coordinates of the actual position on the furnace wall, respectively. u and v are the pixel coordinates in the processed image, respectively. H is the mapping matrix solved using OpenCV functions.

[0074] In practical scenarios, several reference points with definite spatial positions are selected on the heating surface of the furnace to ensure that the boundary of the soot blowing area covered by each soot blower includes at least 8 reference points. The actual reference coordinates (Xt, Yt) of each reference point in the furnace coordinate system are recorded, and its pixel coordinates in the processed image are defined as (Ut, Vt).

[0075] The pixel coordinates (Ut, Vt) in the processed image are determined using the above mapping formula, realizing the mapping and correspondence between (Xt, Yt) and (Ut, Vt). The pixel coordinates of at least 8 reference points included in the boundary of the sootblowing area covered by each sootblower in the processed image are obtained, which is the mapping result. According to the pixel coordinates, the position of the sootblowing area covered by each sootblower in the image is determined. The captured image is divided into several regional images such as region A, region B, and region C. Each region is the image position covered by the sootblower set in the actual furnace. Subsequently, image recognition processing is performed on each regional image.

[0076] It should be noted that since there may be multiple cameras and the image acquisition areas of each camera may overlap, the image that includes the entire blowing area covered by the soot blower can be selected from multiple segmented regional images and used as the regional image corresponding to that soot blower.

[0077] S13. Perform a gray accumulation recognition operation on the region image to obtain the gray accumulation recognition result.

[0078] The dust accumulation identification results include dust accumulation areas and non-dust accumulation areas.

[0079] In one implementation, the region image can be segmented into gray areas and non-gray areas to obtain the gray area recognition result.

[0080] In practice, since the monitoring background of the captured area image is relatively complex, a deep learning image segmentation algorithm with high segmentation accuracy can be used to extract features from the acquired area image, accurately identify and locate the ash accumulation area, quantify the degree of ash accumulation in real time, generate control signals, and provide accurate decision-making basis for subsequent ash blowing operations.

[0081] In one implementation, the deep learning image segmentation algorithm can be the UNet network. That is, the UNet network is used to automatically identify gray areas and non-gray areas (i.e., normal areas).

[0082] Specifically, the UNet network is based on a fully convolutional network (FCN) and can be divided into two parts: an encoder (downsampling path) and a decoder (upsampling path). The encoding part adopts a typical convolutional network architecture, while the decoding part fuses low-level features (feature maps of the encoder) with high-level features (feature maps of the decoder) through deconvolution. This allows for accurate identification, segmentation, and localization of ash accumulation areas. It can better preserve the edges of the soot blowing area and some local details, comprehensively perceive the ash accumulation situation inside the boiler, and provide decision support for subsequent intelligent soot blowing.

[0083] The encoder configuration for the UNet network is as follows:

[0084] The encoder consists of multiple convolutional and pooling layers, which are responsible for extracting image features at different scales.

[0085] ① Convolutional layer

[0086] Convolutional layers are used to extract features from images of the heated surfaces inside the furnace. These features include edges, textures, and shapes, which are crucial for identifying ash-accumulated areas. Through convolutional operations, the model learns local features related to ash accumulation and compresses spatial information, providing a foundation for subsequent classification and segmentation.

[0087] ② Pooling layer

[0088] Pooling operations are used to gradually reduce image resolution, decrease computational cost and feature map size, and enhance high-level features. Common pooling methods include max pooling and average pooling. Max pooling sums the maximum value from each pooling region while retaining the most salient features. Average pooling averages the values ​​from each pooling region. Here, max pooling is used to reduce computational cost while preserving key information related to the gray areas.

[0089] The decoder configuration for the UNet network is as follows:

[0090] The decoder consists of an upsampling layer and a deconvolution layer, which restores the low-resolution feature map to the original image size, enabling the system to intuitively obtain a clear distribution of boiler ash accumulation areas and accurately formulate soot blowing schemes.

[0091] ① Deconvolution layer

[0092] Deconvolutional layers restore the image to high resolution through progressive upsampling, thereby accurately determining the grayscale boundary.

[0093] ② Feature fusion

[0094] Feature fusion refers to merging features from different layers through either concatenation or summation. In this embodiment, concatenation is used to merge the low-resolution feature map from the encoder with the high-resolution feature map from the decoder, achieving more accurate localization and recovery of detailed information.

[0095] The hop connection configuration for the UNet network is as follows:

[0096] Skip connections refer to adding connections between the encoder and decoder, allowing high-resolution features to be directly transmitted to the decoder. This avoids spatial information loss due to oversampling and ensures image processing accuracy. Inside the furnace, ash accumulation patterns are often complex. This approach ensures the algorithm can simultaneously focus on local details and overall ash accumulation trends, thereby improving the detection capability of ash-accumulated areas.

[0097] The activation function configuration for the UNet network is as follows:

[0098] The UNet network employs the ReLU (Rectified Linear Unit) activation function. The ReLU activation function increases the network's non-linearity, prevents gradient vanishing, and allows the model to learn more complex feature representations. In boiler soot blowing control systems, the ReLU function helps the model better capture subtle changes in ash accumulation areas when processing images of the heated surfaces inside the furnace. The activation function formula (ReLU) is as follows:

[0099] f(x) = max(0,x);

[0100] Here, x is the input value, and f(x) is the output value. The ReLU function sets the input value less than 0 to 0, and leaves the input value greater than 0 unchanged.

[0101] In addition, the UNet network can also use the Sigmoid activation function. In the final output, the Sigmoid activation function normalizes the output values ​​to between 0 and 1, directly distinguishing between dust-accumulated and non-dust-accumulated areas, ensuring the accurate execution of subsequent dust-blowing strategies. The formula is:

[0102] ;

[0103] Where x is the input value, This is the output value.

[0104] The loss function configuration for the UNet network is as follows:

[0105] Loss functions measure the difference between model predictions and actual results, and are crucial for training deep learning models. In image segmentation, commonly used loss functions include cross-entropy loss and Dice loss. In boiler soot blowing control systems, ash accumulation areas are typically unevenly distributed. Cross-entropy loss focuses on the classification accuracy of each pixel, easily overlooking small ash accumulation targets, while Dice loss emphasizes the accuracy of the overall segmentation result and is more sensitive to small target detection. Therefore, using cross-entropy loss and Dice loss can better measure the difference between model predictions and actual ash accumulation areas. The formulas for cross-entropy loss and Dice loss are shown below:

[0106] Cross-entropy loss formula: ;

[0107] Dice loss formula: ;

[0108] Total loss function: ;

[0109] Where H is the image height, W is the image width, m and n are the pixel row and column coordinate indices, and y is the image height. m,n For the true label of pixel (m,n) (foreground 1 or background 0), p m,n Let be the predicted foreground probability of pixel (m,n), ε be a small constant to avoid the denominator being zero, q be the weighting coefficient (0~1), and L be the weighting coefficient. CE For cross-entropy loss, L Dice Let L be the Dice loss function, and L be the total loss function.

[0110] By combining these two loss functions, the model can better balance local and global performance during training, thereby improving segmentation accuracy.

[0111] The optimization algorithm for the UNet network employs either SGD (Stochastic Gradient Descent) or Adam (Adaptive Moment Estimation). Specifically, in boiler soot blowing control systems, optimization algorithms such as SGD or Adam are used to update the model's weight parameters to minimize the loss function. Through iterative training, the model gradually learns the ability to identify ash accumulation areas from images of the heated surfaces inside the furnace. The SGD formula is as follows:

[0112] ;

[0113] Where, θ t+1 Let θ be the model parameters updated after the (t+1)th iteration. t Let be the model parameters for the t-th iteration, η be the learning rate, B be the number of mini-batch samples, and s be the sample index. Let be the loss function value for the s-th sample. The gradient with respect to the parameter θ is the biased coefficient of the loss with respect to the parameter.

[0114] In one example, when segmenting gray areas and non-gray areas based on a deep learning image segmentation algorithm, the following specific steps may be included (wherein, this embodiment is only for illustrative purposes, and the specific image size and dataset division can be adjusted according to the actual situation):

[0115] 1. Collect images of the boiler heating surface under different operating conditions and construct a dataset:

[0116] The boiler heating surface video obtained from the industrial camera is extracted frame by frame into a continuous image sequence. The images are preprocessed with distortion correction, grayscale conversion, color enhancement, noise reduction filtering, etc., and the regions are divided according to the soot blower to obtain a 512×512×3 image. The ash accumulation areas in the image are labeled at the pixel level, with ash accumulation areas labeled as 1 and normal areas labeled as 0, and the corresponding single-channel binary mask (label image) is generated.

[0117] It should be noted that the image data in this step must include images of the boiler heating surface under multiple operating conditions such as different coal quality and different loads.

[0118] 2: Dataset partitioning:

[0119] The dataset is divided into three parts: training set, validation set, and test set in a 7:2:1 ratio, and each image data point is guaranteed to exist in only one set.

[0120] It should be noted that the image data in the training set in this step cannot be images of the boiler heating surface under the same operating conditions.

[0121] 3: UNet model setup:

[0122] 3.1 Input Layer

[0123] The preprocessed 512×512×3 image is input into the model.

[0124] 3.2 Encoder (Downsampling Path)

[0125] The encoder employs four modules, each using two convolutions and one pooling to progressively reduce resolution and extract more image features. Simultaneously, the feature maps before each pooling are saved and subsequently concatenated with the corresponding layers of the decoder.

[0126] Module 1: First, two 3×3 convolutions are performed. The first convolution expands the image channels to 64, outputting a 512×512×64 image. The second convolution maintains 64 channels, enhances feature extraction, and outputs another 512×512×64 image. Then, max pooling is performed using a 2×2 pooling window to compress the image size to 256×256×64, and this image is input into Module 2. Simultaneously, the feature map before pooling (512×512×64) is saved.

[0127] Module 2: First, two 3×3 convolutions are performed. The first convolution increases the image channels to 128, outputting a 256×256×128 image. The second convolution maintains 128 channels, enhances feature extraction, and outputs another 256×256×128 image. Then, max pooling is performed using a 2×2 pooling window to compress the image size to 128×128×128, and this image is input into Module 3. Simultaneously, the feature map before pooling (256×256×128) is saved.

[0128] Module 3: First, two 3×3 convolutions are performed. The first convolution increases the image channels to 256, outputting a 128×128×256 image. The second convolution maintains the 256 channels, enhances feature extraction, and outputs another 128×128×256 image. Then, max pooling is performed using a 2×2 pooling window to compress the image size to 64×64×256, and this image is input into Module 4. Simultaneously, the feature map before pooling (128×128×256) is saved.

[0129] Module 4: First, two 3×3 convolutions are performed. The first convolution increases the image channels to 512, outputting a 64×64×512 image. The second convolution maintains the 512 channels, enhances feature extraction, and outputs another 64×64×512 image. Then, max pooling is performed using a 2×2 pooling window to compress the image size to 32×32×512. Simultaneously, the feature map before pooling (64×64×512) is saved.

[0130] Then, the lowest-level features are extracted using Bottlenck, and after two convolutions, a 32×32×1024 image with double the number of channels is obtained, which is then input into the decoder.

[0131] 3.3 Decoder (Upsampling Path)

[0132] The decoder employs four modules, each of which gradually restores resolution through two convolutions and one pooling. Simultaneously, it stitches the images from the corresponding modules in the encoder along the channel dimension using skip connections.

[0133] Module 5: First, deconvolve using a 2×2 convolution kernel to obtain a 64×64×512 image. Then, concatenate this image with the 64×64×512 feature map stored in Module 4 to obtain a 64×64×1024 image. Finally, perform two convolutions to obtain a 64×64×512 image and transmit the image to Module 6.

[0134] Module Six: First, deconvolve using a 2×2 convolution kernel to obtain a 128×128×256 image. Then, concatenate this image with the 128×128×256 feature map saved in Module Three to obtain a 128×128×512 image. Finally, perform two convolutions to obtain a 128×128×256 image, and then transmit the image to Module Seven.

[0135] Module 7: First, deconvolve using a 2×2 convolution kernel to obtain a 256×256×128 image. Then, concatenate this image with the 256×256×128 feature map saved in Module 2 to obtain a 256×256×256 image. Finally, perform two convolutions to obtain a 256×256×128 image and transmit the image to Module 8.

[0136] Module 8: First, deconvolve using a 2×2 convolution kernel to obtain a 512×512×64 image. Then, concatenate this image with the 512×512×64 feature map saved in Module 1 to obtain a 512×512×128 image. Finally, convolve twice to obtain a 512×512×64 image and transmit the image to the output layer.

[0137] 3.4 Output Layer

[0138] The output layer convolves the image using a 1×1 kernel, outputting a 512×512×1 segmentation map. Each pixel is assigned a probability value between 0 and 1, representing the likelihood that it belongs to a gray area. By setting a probability threshold of 0.5, areas above the threshold are classified as "gray areas" and labeled as 1, while areas below the threshold are classified as "non-gray areas (i.e., normal areas)" and labeled as 0. The specific threshold value can be dynamically adjusted by engineers to improve the accuracy and practicality of gray area recognition.

[0139] 4: Image segmentation and feature parameter extraction:

[0140] Specifically, the real-time boiler heating surface video captured by the image acquisition module is extracted frame by frame into a continuous image sequence, preprocessed and mapped and segmented, and then input into the model. The UNet neural network processes the image to automatically identify the ash accumulation area and the non-ash accumulation area, and obtain the ash accumulation identification result.

[0141] S14. Based on the ash accumulation recognition results, calculate the regional attribute information of the region image.

[0142] The regional attribute information includes regional dust coverage, average dust thickness, and regional roughness.

[0143] Specifically, after obtaining the dust accumulation recognition results, image feature values ​​such as regional dust accumulation coverage rate and regional average dust accumulation thickness can be calculated.

[0144] In one implementation, such as Figure 2 As shown, step S14 may include:

[0145] S21. Calculate the proportion of pixels located in the gray area to the total number of pixels in the region image to obtain the region gray coverage rate.

[0146] Specifically, after segmenting the region image into gray areas, the region gray coverage rate is calculated. The region gray coverage rate refers to the proportion of pixels in the gray area to the total number of pixels in the region image, reflecting the size of the gray area. The calculation formula is as follows:

[0147] ;

[0148] Where Ci is the area dust coverage rate of the region image of the i-th soot blower. This represents the number of pixels with gray in that region of the image. This represents the total number of pixels in the image for that region.

[0149] S22. Calculate the average gray value of the region image to obtain the average gray thickness of the region.

[0150] Specifically, the average ash thickness of a region is obtained by calculating the average grayscale value of the region image, and is used to characterize the average thickness of the ash layer within the region. The average grayscale value of the region image under clean boiler conditions is defined as the baseline value I. base The formula for calculating the average ash accumulation thickness in a region is:

[0151] ;

[0152] in, Let I be the average dust accumulation thickness in the i-th sootblower region, and let I(x,y) be the grayscale value corresponding to the coordinate (x,y). This represents the total number of pixels in the region. To represent a region in a region image.

[0153] S23. Calculate the regional texture uniformity, regional grayscale difference, and regional texture complexity of the region image.

[0154] Specifically, the real-time roughness L of a region reflects the irregularity and undulation of the texture on the heated surface caused by ash accumulation. It consists of three parts: region texture uniformity U, region grayscale difference C, and region texture complexity E. When ash accumulates on the heated surface of the furnace water-cooled wall, the originally regular structure will exhibit blurred boundaries and irregular distribution, resulting in lower region texture uniformity, higher contrast, higher region texture complexity, and increased roughness. The calculation formula is:

[0155] ;

[0156] ;

[0157] in, , , These are the weighting coefficients, The normalized region texture uniformity, This represents the normalized regional grayscale difference. The normalized region texture complexity, This represents the real-time roughness of the region.

[0158] The calculation process is as follows:

[0159] First calculate The calculation formula is as follows:

[0160] ;

[0161] in, Let R represent the probability of the gray-level pair (h, l) appearing in the entire region image, where R represents the region in the region image, h is the quantized gray-level value of the first pixel, and l is the quantized gray-level value of the adjacent pixel. h and l iterate through all possible combinations of gray-level values. This represents the number of times a pixel with gray level h appears in R along with its adjacent pixel with gray level l (hereinafter, a pixel with gray level h and its adjacent pixel with gray level j are called a gray level pair). G represents the sum of the occurrences of all gray-level pairs formed by all pixels and their adjacent pixels. G is the total number of gray levels, which represents the number of levels after discretizing the gray-level values ​​of the image. It is usually taken as 16 to 32 levels in order to reduce the dimension of the gray-level co-occurrence matrix and improve the efficiency of feature calculation.

[0162] The values ​​of h and l are both quantized grayscale values. The grayscale quantization discretization formula is as follows:

[0163] ;

[0164] in, G represents the quantized grayscale value, where G is the total number of grayscale levels and g is the original grayscale value (e.g., 0~255). This indicates rounding down. This is the highest value of the original grayscale.

[0165] In the calculation After that, it can be calculated that The specific calculation formula is as follows:

[0166] ;

[0167] ;

[0168] Where U represents the uniformity of the region texture. This is the normalized region texture uniformity, used to measure the uniformity of image grayscale distribution. If the image texture distribution is more regular, there are fewer grayscale areas, and certain grayscale pairs have a higher probability of appearing. They will gather together. The larger the value, the better. If the image texture is complex and the grayscale distribution is relatively dispersed, then... The value is small.

[0169] The calculation process is as follows:

[0170] ;

[0171] ;

[0172] Where C represents the regional grayscale difference. This is the normalized regional grayscale difference, used to describe the intensity of the difference in grayscale values ​​between adjacent pixels. The larger the value, the more significant the grayscale difference between adjacent regions, and the more turbulent the surface grayscale layer; the smaller the value, the smoother the surface grayscale layer.

[0173] The calculation process is as follows:

[0174] ;

[0175] ;

[0176] E represents the region texture complexity. This is the normalized region texture complexity, used to describe the uncertainty and randomness of image texture. The larger the value, the higher the texture complexity and the more uneven the dust accumulation; the smaller the value, the simpler the texture and the more regular the dust accumulation layer.

[0177] S24. Calculate the real-time roughness of the region image based on the region texture uniformity, region grayscale difference, and region texture complexity.

[0178] Specifically, by substituting the region texture uniformity, region grayscale difference, and region texture complexity into the calculation formula for the region's real-time roughness L, the region's real-time roughness of the image can be obtained.

[0179] S25. Calculate the regional dust accumulation roughness of the regional image based on the real-time regional roughness and the reference roughness.

[0180] Specifically, considering that the heating surface of the furnace is composed of water-cooled wall tubes and tube spacers, and that striations also exist before ash accumulation, before calculating the roughness, images of the heating surface are collected in the boiler shutdown maintenance or clean state and the roughness in this state is calculated and defined as the reference roughness Lbase.

[0181] Therefore, the roughness of the area of ​​dust accumulation is defined as follows: The calculation formula is as follows:

[0182] ;

[0183] Where L is the real-time roughness of the region, L base The reference roughness.

[0184] In this embodiment, the real-time ash accumulation situation is fully obtained from three perspectives: regional texture uniformity, regional grayscale difference, and regional texture complexity of the furnace heating surface. Subsequently, the regional ash accumulation coverage, regional average ash accumulation thickness, and regional ash accumulation roughness are used as inputs to the fuzzy control model.

[0185] S15. Determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function.

[0186] In this embodiment, the intelligent control module combines the boiler operating conditions with changes in ash accumulation in the image, and uses a fuzzy control algorithm to dynamically adjust the parameters and actions of the soot blowing execution module, thereby achieving precise control of the soot blowing process. First, the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information are determined. Then, an adaptive fuzzy membership function is constructed using these boundary values. Finally, the fuzzy quantity corresponding to the regional attribute information is determined using the adaptive fuzzy membership function.

[0187] S16. Based on the fuzzy amount corresponding to the regional attribute information and the fuzzy rule library, determine the blowing parameters of the soot blower corresponding to the regional image, and control the corresponding soot blower to perform the blowing operation according to the blowing parameters.

[0188] Specifically, through the aforementioned fuzzification and subsequent fuzzy inference and defuzzification operations, the blowing parameters of the soot blower corresponding to the region image are determined based on the fuzziness amount corresponding to the region attribute information and the fuzzy rule base.

[0189] The soot blowing parameters can be at least the soot blowing intensity coefficient and the start / stop command. The soot blowing operation is then performed by controlling the corresponding soot blower according to these parameters. Specifically, when the start / stop command is "start," the soot blowing operation is performed according to the soot blowing intensity coefficient.

[0190] In this embodiment, an image acquisition operation is performed on the boiler heating surface to obtain an initial image. The initial image undergoes image processing such as distortion correction, grayscale conversion, color enhancement, and noise reduction filtering to obtain a processed image. The pixel coordinates of the processed image are mapped to the actual reference coordinates of the boiler heating surface to obtain a mapping result. Based on the mapping result, the processed image is segmented according to the sootblowing area covered by each sootblower to obtain the region image corresponding to the sootblower. Ash accumulation recognition is performed on the region image to obtain the ash accumulation recognition result. The ash accumulation recognition result includes ash-accumulated areas and non-ash-accumulated areas. Based on the ash accumulation recognition result, the region attribute information of the region image is calculated. The region attribute information includes ash accumulation coverage, average ash accumulation thickness, and region roughness. The boundary values ​​of the fuzzy membership function corresponding to the region attribute information are determined. An adaptive fuzzy membership function is constructed using the boundary values ​​of the fuzzy membership function. The fuzzy quantity corresponding to the region attribute information is determined using the adaptive fuzzy membership function. Based on the fuzzy quantity corresponding to the region attribute information and the fuzzy rule base, the sootblowing parameters of the sootblower corresponding to the region image are determined. The sootblower is controlled to perform sootblowing operations according to the sootblowing parameters. In this application, the ash accumulation is analyzed based on the real-time operating data of the boiler heating surface to obtain accurate soot blowing parameters, thereby enabling precise control of boiler soot blowing.

[0191] Based on any of the above embodiments, refer to Figure 3 Determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, including:

[0192] S31. Perform clustering operations on the regional attribute information to obtain initial boundary values.

[0193] The initial boundary value refers to the initial boundary value of the fuzzy membership function.

[0194] In this embodiment, the regional attribute information includes regional dust coverage, average regional dust thickness, and regional roughness. Since there are multiple regional images, the regional dust coverage, average regional dust thickness, and regional roughness corresponding to each regional image can be obtained. For each of the regional dust coverage, average regional dust thickness, and regional roughness, a clustering operation is performed on the corresponding parameter for all regional images. The clustering algorithm can be the K-means clustering algorithm, with the number of categories K=3, and the categories divided into "low / medium / high". Taking regional dust coverage as an example, the regional dust coverage of all regional images is clustered to obtain the cluster centers of the three major categories corresponding to the regional dust coverage: high, medium, and low. The three cluster centers are used as the center points a of the three major categories high, medium, and low, respectively. The left boundary b and right boundary c of the three major categories high, medium, and low are determined according to the nearest neighbor mean method. At this time, the center point a, left boundary b, and right boundary c of each of the three major categories high, medium, and low are obtained. The center point a, left boundary b, and right boundary c can be used as the initial boundary values.

[0195] S32. Based on the historical soot blowing data of the soot blower, obtain the correction value.

[0196] In this embodiment, when using fuzzy control for soot blowing control, to ensure the adaptability of fuzzy control decoupling, a dynamic adjustment mechanism is used to correct the (a, b, c) parameters in real time. The dynamic adjustment mapping function is defined as follows:

[0197] ;

[0198] ;

[0199] in, For the real-time values ​​of input variables (such as region attribute information), This refers to the actual soot blowing effect feedback index (such as soot blowing intensity), where 's' represents three categories: high, medium, and low. These represent the left boundary, center point, and right boundary of the three categories: high, medium, and low, respectively, with specific values ​​a, b, and c as mentioned above. The adjustment coefficient is set according to the actual working conditions. To provide feedback metrics on the desired dust removal effect, in real-world scenarios, if it is found that using actual dust removal effect feedback metrics... When performing soot blowing control, if the soot blowing effect is unsatisfactory, feedback indicators for the expected soot blowing effect can be set based on experience. . For function, These are the weighting coefficients. The input is normalized weights, so that the adjustment amount varies with the input size.

[0200] The above functions enable dynamic adjustment when ash accumulation characteristics or operating conditions change. This adapts to changes, ensuring the accuracy and adaptability of fuzzy control.

[0201] In one implementation, feedback indicators of the actual sootblowing effect of the sootblower during the previous sootblowing operation can be obtained. and actual soot blowing effect feedback indicators Corresponding expected dust removal effect feedback indicators Obtain the minimum area attribute information of the sootblower during historical sootblowing operations (i.e. ) and minimum region attribute information (i.e. ).

[0202] Finally, based on the differences between the actual and expected soot blowing effect feedback indicators, the differences between the area attribute information and the minimum area attribute information, and the differences between the maximum and minimum area attribute information, correction values ​​are calculated. Specifically, these correction values ​​are calculated using the aforementioned dynamic adjustment mapping function. , as well as These three parameters are called correction values.

[0203] S33. Use the sum of the initial boundary value and the correction value as the boundary value of the fuzzy membership function.

[0204] Specifically, the boundary values ​​of the fuzzy membership function are calculated using the aforementioned dynamically adjusted mapping function, including: the left boundary. Center point and right boundary .

[0205] In this embodiment, an adaptive adjustment algorithm is used to update the boundary values ​​of the fuzzy membership function online to avoid the uncertainty caused by manual division, while also adapting to the differences under different boilers and different coal-fired operating conditions.

[0206] Furthermore, by introducing self-learning and optimization mechanisms, the system can automatically adjust control strategies and algorithm parameters based on historical data and real-time feedback, improving the system's adaptability and control effectiveness. In addition, a user-friendly interface provides a convenient way for operators to monitor the system's operating status in real time and make necessary adjustments and optimizations.

[0207] Based on the above embodiments, one implementation method, which uses an adaptive fuzzy membership function to determine the fuzzy quantity corresponding to the regional attribute information, may include:

[0208] The region attribute information is fuzzified using an adaptive fuzzy membership function to obtain the fuzzy quantity corresponding to the region attribute information.

[0209] Specifically, a membership function is used to partition each input variable, employing a triangular membership function to convert the actual value into a fuzzy quantity. The formula for the adaptive fuzzy membership function is as follows:

[0210] ;

[0211] in, This represents the membership degree of the input variable X to the fuzzy set, i.e., the fuzzy quantity, where s represents three categories: high, medium, and low. , and These are the fuzzy membership function boundary values ​​corresponding to the left boundary, center point, and right boundary of the category (high / medium / low), respectively.

[0212] By inputting the regional attribute information into the adaptive fuzzy membership function, the fuzzy quantity corresponding to the regional attribute information can be obtained.

[0213] It should be noted that the intelligent control module can also optionally use boiler operating parameters as input variables X in the fuzzy input. In this case, input variable X includes regional attribute information and boiler operating parameters, including but not limited to flue gas temperature deviation, oxygen deviation, pressure difference fluctuation and unit load.

[0214] Based on this embodiment, in one implementation, refer to Figure 4 Based on the fuzziness amount corresponding to the regional attribute information and the fuzzy rule base, the blowing parameters of the soot blower corresponding to the regional image are determined, including:

[0215] S41. Obtain the pre-established fuzzy rule base.

[0216] In this embodiment, the fuzzy rule base can be configured according to actual needs. In one example, a fuzzy rule base is established based on expert experience and experimental data; the specific fuzzy rule base is as follows:

[0217] If the area has a high ash coverage rate, a high average ash thickness, and a high boiler load, then the soot blowing intensity is high.

[0218] If the area has a medium ash coverage rate, a medium average ash thickness, and a high boiler load, then the soot blowing intensity is medium.

[0219] If the area has a medium dust accumulation coverage rate, a low average dust accumulation thickness, and a medium roughness, then the dust blowing intensity is medium.

[0220] If the area has low dust accumulation coverage, low average dust accumulation thickness, and low roughness, then the dust blowing intensity is low.

[0221] In this embodiment, the fuzzy rule base is stored in the form of a two-dimensional or three-dimensional table, which supports real-time modification. The specific number and details of the rules are set according to the actual blowing area and operating characteristics.

[0222] S42. Calculate the rule activation degree based on the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, perform fuzzy inference based on the rule activation degree, and obtain the output fuzzy quantity.

[0223] Specifically, in the calculation Subsequently, based on the set fuzzy rule base, a Mamdani-type inference mechanism is adopted, and the maximum-minimum synthesis method is used to realize the matching and output inference of fuzzy sets, and the output fuzzy quantity is obtained.

[0224] Specifically, firstly, based on the fuzzy quantities corresponding to the regional attribute information and the fuzzy rule base, the activation degree of a single rule is calculated using the minimum method, and then the inference results of multiple rules are synthesized using the maximum method to obtain the output fuzzy quantity.

[0225] S43. Using the output blur quantity, perform deblurring operation through the centroid method to obtain the soot blowing parameters of the soot blower corresponding to the region image.

[0226] Among them, the soot blowing parameters include at least the soot blowing intensity coefficient and the start and stop commands.

[0227] In this embodiment, the centroid method is used to deblur the output fuzzy value to obtain the final output value, i.e. Values. The output values ​​include at least the start / stop command S of the sootblower and the sootblowing intensity P, calculated using the following formula:

[0228] ;

[0229] in, The final output clarity values ​​(such as the soot blowing intensity coefficient P and the start / stop command S), where s represents three categories: high, medium, and low. The output fuzzy membership degree of s (the result of multiple rules superimposed). The output values ​​are for high, medium, and low (e.g., low = 1, medium = 5, high = 7).

[0230] The start / stop command S and the soot blowing intensity coefficient P are transmitted to the execution module to achieve intelligent soot blowing control. Simultaneously, parameters are automatically adjusted based on real-time data changes to achieve optimized control of the soot blowing system.

[0231] In practice, the input signal of the intelligent control module is directly transmitted to the soot blowing actuator (such as a soot blowing actuator), and after soot blowing, the control strategy and algorithm parameters are automatically adjusted through real-time image feedback and data changes to achieve closed-loop optimization control of the soot blowing system, thereby improving the system's adaptability and control accuracy.

[0232] Unlike traditional fuzzy control methods that rely on operating parameters and human experience, this invention introduces multi-source fusion input of image recognition features and boiler operating parameters into the intelligent control module. It dynamically identifies the distribution characteristics of input variables through clustering algorithms, adaptively adjusts the boundary values ​​of fuzzy membership functions, and combines closed-loop feedback optimization to achieve intelligent, refined, and adaptive control of the sootblower, significantly improving the accuracy of ash accumulation discrimination and the rationality of sootblowing decisions.

[0233] In summary, the intelligent soot blowing control system for boilers in this application fully integrates image recognition technology with intelligent control algorithms, enabling real-time monitoring and precise control of the soot blowing situation on the boiler's heating surface. This system not only improves the boiler's operating efficiency and heat exchange efficiency but also reduces energy consumption and maintenance costs, providing a strong guarantee for the safe and efficient operation of the boiler.

[0234] like Figure 5 As shown, the key technical points of the intelligent soot blowing control system for boilers of the present invention mainly include:

[0235] (I) Optimization of Image Acquisition and Preprocessing Techniques

[0236] High-resolution, wide-angle, and high-temperature resistant cameras are installed at key observation locations on the boiler to ensure comprehensive and clear image capture of the heated surfaces.

[0237] Advanced image processing techniques (distortion correction, image grayscale conversion, color enhancement, noise reduction filtering, image mapping, etc.) are employed to improve the stability and anti-interference ability of image signals. At the same time, the region is mapped to the area where the corresponding soot blower is applied, providing high-quality image data for subsequent processing.

[0238] (II) Image Processing Algorithms Based on Deep Learning

[0239] We developed and used a deep learning image segmentation algorithm (UNet network) for image segmentation and recognition, accurately identifying ash accumulation areas and quantifying the ash accumulation status of boiler heating systems. Refer to the corresponding description above for the specific structure.

[0240] Optimize algorithm parameters to improve the algorithm's processing speed and accuracy, and ensure the real-time performance and reliability of image processing results.

[0241] (III) Extraction of image feature parameters

[0242] Based on real-time images, core feature parameters such as regional ash coverage, average regional ash thickness, and regional roughness are extracted. Combined with the difference between the baseline value in the clean state and the real-time value in the ash-covered state, the ash state of the heated area is comprehensively characterized, forming a quantifiable control input quantity, which provides data support for intelligent soot blowing.

[0243] (iv) Design of intelligent control strategies and algorithms

[0244] K-means clustering algorithm is used to divide intervals, and fuzzification processing is performed using adaptive triangular membership function to establish a fuzzy rule base for ash accumulation characteristics and soot blowing intensity. Multi-parameter comprehensive inference is achieved based on rule activation degree. Simultaneously, by combining real-time boiler operating conditions and process parameters, soot blowing command parameters are adaptively adjusted to achieve dynamic optimization of soot blowing intensity and frequency, avoiding over- or under-blowing, improving system adaptability and control accuracy, and ensuring the cleanliness and economy of boiler operation.

[0245] The key technologies of the intelligent soot blowing control system for boilers in this application lie in the optimization of image acquisition and preprocessing technologies, deep learning-based image segmentation technology, extraction of image feature parameters, design of intelligent control strategies and algorithms, and the implementation of self-learning and optimization functions. These together form the foundation for the system's efficient, intelligent, and reliable operation, providing strong support for the safe and efficient operation of the boiler. Compared with existing technologies, this technical solution has the following characteristics:

[0246] (a) Improve boiler operating efficiency and heat exchange efficiency

[0247] By monitoring the ash accumulation on the boiler's heating surfaces in real time and precisely controlling the soot blowing process, the ash accumulation on the heating surfaces is effectively reduced, thereby improving the boiler's heat transfer efficiency and heat exchange efficiency.

[0248] (ii) Reduce energy consumption and maintenance costs

[0249] This application achieves precise control of the soot blowing process through an intelligent control strategy, avoiding unnecessary soot blowing operations and thus reducing energy consumption. Simultaneously, by reducing ash accumulation on the heating surfaces, it also extends the boiler's service life and lowers maintenance costs.

[0250] (III) Improve the reliability of automatic control

[0251] The system of this application employs advanced image recognition technology and intelligent control algorithms to achieve real-time monitoring and precise control of ash accumulation on the boiler's heating surface. Compared with existing manual control or simple timed control, the system of this application has higher reliability and stability in automatic control, ensuring the safe and efficient operation of the boiler.

[0252] (iv) Reduce the need for manual intervention

[0253] The system invented in this application enables intelligent control of the boiler soot blowing process, greatly reducing the need for manual intervention. Operators can remotely manage the entire system by simply monitoring and making necessary adjustments through a human-machine interface. This not only reduces labor intensity but also improves work efficiency.

[0254] (v) It is conducive to environmental protection and energy conservation and emission reduction.

[0255] By improving the boiler's operating efficiency and heat exchange efficiency, the system of this application helps reduce the consumption of fossil fuels such as coal, thereby reducing emissions of greenhouse gases such as carbon dioxide. Simultaneously, by reducing ash accumulation on the heating surfaces, it also reduces pollutant emissions during boiler operation, contributing to environmental protection and energy conservation.

[0256] The intelligent soot blowing control system for boilers proposed in this application has significant advantages in improving boiler operating efficiency and heat exchange efficiency, reducing energy consumption and maintenance costs, improving the reliability of automatic control, reducing the need for manual intervention, and promoting environmental protection, energy conservation and emission reduction.

[0257] Based on the embodiments of the above-described boiler soot blowing control method, another embodiment of this application provides a boiler soot blowing control device, referring to... Figure 6 ,include:

[0258] Image acquisition module 101 is used to perform image acquisition operations on the boiler heating surface to obtain an initial image;

[0259] The image processing module 102 is used to perform image processing operations on the initial image to obtain the processed image, map the pixel coordinates in the processed image to the actual reference coordinates of the boiler heating surface to obtain the mapping result, and based on the mapping result, perform image segmentation operations on the processed image according to the soot blowing area covered by each soot blower to obtain the region image corresponding to the soot blower; the image processing operations include at least distortion correction, grayscale conversion, color enhancement and noise reduction filtering;

[0260] The ash accumulation recognition module 103 is used to perform ash accumulation recognition on a region image to obtain ash accumulation recognition results; the ash accumulation recognition results include ash accumulation regions and non-ash accumulation regions.

[0261] The information calculation module 104 is used to calculate the regional attribute information of the region image based on the ash accumulation recognition result; the regional attribute information includes the region ash accumulation coverage, the region average ash accumulation thickness, and the region roughness.

[0262] The fuzzy quantity determination module 105 is used to determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function.

[0263] The soot blowing control module 106 is used to determine the soot blowing parameters of the soot blower corresponding to the region image based on the fuzziness amount corresponding to the region attribute information and the fuzzy rule library, and control the corresponding soot blower to perform soot blowing operation according to the soot blowing parameters.

[0264] In one implementation, the ash accumulation recognition module 103 is specifically used for:

[0265] The image is segmented into gray areas and non-gray areas to obtain the gray area recognition result.

[0266] In one implementation, the information computing module 104 includes:

[0267] The first calculation submodule is used to calculate the proportion of pixels located in the gray area to the total number of pixels in the region image, and obtain the region gray coverage rate.

[0268] The second calculation submodule is used to calculate the average gray value of the region image to obtain the average gray thickness of the region.

[0269] The third calculation submodule is used to calculate the regional texture uniformity, regional grayscale difference, and regional texture complexity of the regional image.

[0270] The fourth calculation submodule is used to calculate the real-time roughness of the region image based on the region texture uniformity, region grayscale difference and region texture complexity.

[0271] The fifth calculation submodule is used to calculate the regional dust accumulation roughness of the region image based on the real-time regional roughness and the reference roughness.

[0272] In one implementation, the fuzzy quantity determination module 105 includes:

[0273] The clustering submodule is used to perform clustering operations on region attribute information to obtain initial boundary values;

[0274] The correction value determination submodule is used to obtain correction values ​​based on historical soot blowing data from the soot blower.

[0275] The boundary value determination submodule is used to use the sum of the initial boundary value and the correction value as the boundary value of the fuzzy membership function.

[0276] In one implementation, the correction value determination submodule includes:

[0277] The acquisition unit is used to acquire the actual soot blowing effect feedback index of the soot blower during the last soot blowing operation and the expected soot blowing effect feedback index corresponding to the actual soot blowing effect feedback index; and to acquire the minimum area attribute information and maximum area attribute information of the soot blower during historical soot blowing operations.

[0278] The calculation unit is used to calculate correction values ​​based on the difference between the actual soot blowing effect feedback index and the expected soot blowing effect feedback index, the difference between the regional attribute information and the minimum regional attribute information, and the difference between the maximum regional attribute information and the minimum regional attribute information.

[0279] In one implementation, the fuzzy quantity determination module 105 includes:

[0280] The fuzzy processing submodule is used to perform fuzzification operations on regional attribute information using an adaptive fuzzy membership function to obtain the fuzzy quantity corresponding to the regional attribute information.

[0281] In one implementation, the soot blowing control module 106 includes:

[0282] The rule base acquisition submodule is used to acquire a pre-established fuzzy rule base;

[0283] The fuzzy quantity calculation submodule is used to calculate the rule activation degree based on the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, perform fuzzy inference based on the rule activation degree, and obtain the output fuzzy quantity.

[0284] The parameter determination submodule is used to perform deblurring operation by using the centroid method to obtain the soot blowing parameters of the soot blower corresponding to the region image using the output fuzzy quantity; the soot blowing parameters include at least the soot blowing intensity coefficient and start / stop commands.

[0285] In this embodiment, an image acquisition operation is performed on the boiler heating surface to obtain an initial image. The initial image undergoes image processing such as distortion correction, grayscale conversion, color enhancement, and noise reduction filtering to obtain a processed image. The pixel coordinates of the processed image are mapped to the actual reference coordinates of the boiler heating surface to obtain a mapping result. Based on the mapping result, the processed image is segmented according to the sootblowing area covered by each sootblower to obtain the region image corresponding to the sootblower. Ash accumulation recognition is performed on the region image to obtain the ash accumulation recognition result. The ash accumulation recognition result includes ash-accumulated areas and non-ash-accumulated areas. Based on the ash accumulation recognition result, the region attribute information of the region image is calculated. The region attribute information includes ash accumulation coverage, average ash accumulation thickness, and region roughness. The boundary values ​​of the fuzzy membership function corresponding to the region attribute information are determined. An adaptive fuzzy membership function is constructed using the boundary values ​​of the fuzzy membership function. The fuzzy quantity corresponding to the region attribute information is determined using the adaptive fuzzy membership function. Based on the fuzzy quantity corresponding to the region attribute information and the fuzzy rule base, the sootblowing parameters of the sootblower corresponding to the region image are determined. The sootblower is controlled to perform sootblowing operations according to the sootblowing parameters. In this application, the ash accumulation is analyzed based on the real-time operating data of the boiler heating surface to obtain accurate soot blowing parameters, thereby enabling precise control of boiler soot blowing.

[0286] It should be noted that the working process of each module, submodule and unit in this embodiment is described in the corresponding description in the above embodiment, and will not be repeated here.

[0287] This application also provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0288] Memory is used to store computer programs;

[0289] The processor is used to execute computer programs so that the electronic equipment can implement the above-described boiler soot blowing control method.

[0290] refer to Figure 7 As shown, Figure 7 A schematic diagram of the structure of an electronic device according to an embodiment of this application is shown. The electronic device in the embodiment of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 7 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0291] like Figure 7 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0292] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, memory cards, hard drives, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0293] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the boiler soot blowing control methods provided in this application.

[0294] This application also provides a computer storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the boiler soot blowing control methods provided in this application.

[0295] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0296] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0297] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0298] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A boiler soot blowing control method, characterized in that, include: An image acquisition operation is performed on the boiler heating surface to obtain an initial image; The initial image is processed to obtain a processed image. The pixel coordinates in the processed image are mapped to the actual reference coordinates of the boiler heating surface to obtain a mapping result. Based on the mapping result, the processed image is segmented according to the sootblowing area covered by each sootblower to obtain the region image corresponding to the sootblower. The image processing operation includes at least distortion correction, grayscale conversion, color enhancement, and noise reduction filtering. A gray accumulation recognition operation is performed on the image of the region to obtain a gray accumulation recognition result; the gray accumulation recognition result includes gray accumulation regions and non-gray accumulation regions. Based on the dust accumulation recognition result, calculate the regional attribute information of the region image; the regional attribute information includes the region dust accumulation coverage, the region average dust accumulation thickness, and the region roughness. Determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function; Based on the fuzziness amount corresponding to the regional attribute information and the fuzzy rule library, the blowing parameters of the soot blower corresponding to the regional image are determined, and the soot blower is controlled to perform the blowing operation according to the blowing parameters.

2. The boiler soot blowing control method according to claim 1, characterized in that, Perform a gray accumulation recognition operation on the image of the region to obtain the gray accumulation recognition result, including: The image of the region is segmented into gray areas and non-gray areas to obtain the gray area recognition result.

3. The boiler soot blowing control method according to claim 1, characterized in that, Based on the ash accumulation recognition result, the regional attribute information of the region image is calculated, including: Calculate the proportion of pixels located in the gray area to the total number of pixels in the region image to obtain the region gray coverage rate; Calculate the average gray value of the region image to obtain the average gray thickness of the region. Calculate the region texture uniformity, region grayscale difference, and region texture complexity of the region image; The real-time roughness of the region image is calculated based on the region texture uniformity, the region grayscale difference, and the region texture complexity. The regional roughness of the image is calculated based on the real-time roughness of the region and the reference roughness.

4. The boiler soot blowing control method according to claim 1, characterized in that, Determining the fuzzy membership function boundary values ​​corresponding to the region attribute information includes: Clustering operations are performed on the region attribute information to obtain initial boundary values; Based on the historical soot blowing data of the soot blower, a correction value is obtained; The sum of the initial boundary value and the correction value is used as the boundary value of the fuzzy membership function.

5. The boiler soot blowing control method according to claim 4, characterized in that, Based on the historical soot blowing data of the soot blower, correction values ​​are obtained, including: Obtain the actual soot blowing effect feedback index of the soot blower during the previous soot blowing operation and the expected soot blowing effect feedback index corresponding to the actual soot blowing effect feedback index. Obtain the minimum and maximum area attribute information of the sootblower during historical sootblowing operations; Based on the difference between the actual soot blowing effect feedback index and the expected soot blowing effect feedback index, the difference between the regional attribute information and the minimum regional attribute information, and the difference between the maximum regional attribute information and the minimum regional attribute information, a correction value is calculated.

6. The boiler soot blowing control method according to claim 1, characterized in that, Determining the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function includes: The region attribute information is fuzzified using the adaptive fuzzy membership function to obtain the fuzzy quantity corresponding to the region attribute information.

7. The boiler soot blowing control method according to claim 1, characterized in that, Based on the blur amount corresponding to the region attribute information and the blur rule library, the blowing parameters of the soot blower corresponding to the region image are determined, including: Obtain a pre-established fuzzy rule base; Based on the fuzzy quantity corresponding to the regional attribute information and the fuzzy rule base, the rule activation degree is calculated, and fuzzy inference is performed based on the rule activation degree to obtain the output fuzzy quantity. Using the output blur quantity, a deblurring operation is performed using the centroid method to obtain the blowing parameters of the soot blower corresponding to the region image; the blowing parameters include at least the blowing intensity coefficient and start / stop commands.

8. A boiler soot blowing control device, characterized in that, include: The image acquisition module is used to acquire images of the boiler's heating surface to obtain initial images. The image processing module is used to perform image processing operations on the initial image to obtain a processed image, map the pixel coordinates in the processed image to the actual reference coordinates of the boiler heating surface to obtain a mapping result, and based on the mapping result, perform image segmentation operations on the processed image according to the soot blowing area covered by each soot blower to obtain the region image corresponding to the soot blower; the image processing operations include at least distortion correction, grayscale conversion, color enhancement, and noise reduction filtering; The dust accumulation recognition module is used to perform dust accumulation recognition on the region image to obtain a dust accumulation recognition result; the dust accumulation recognition result includes a dust accumulation region and a non-dust accumulation region. The information calculation module is used to calculate the regional attribute information of the region image based on the dust accumulation recognition result; the regional attribute information includes the region dust accumulation coverage, the region average dust accumulation thickness, and the region roughness. The fuzzy quantity determination module is used to determine the boundary values ​​of the fuzzy membership function corresponding to the regional attribute information, construct an adaptive fuzzy membership function using the boundary values ​​of the fuzzy membership function, and determine the fuzzy quantity corresponding to the regional attribute information using the adaptive fuzzy membership function. The soot blowing control module is used to determine the soot blowing parameters of the soot blower corresponding to the region image based on the fuzziness amount corresponding to the region attribute information and the fuzzy rule library, and control the corresponding soot blower to perform soot blowing operation according to the soot blowing parameters.

9. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the boiler soot blowing control method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The computer storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the boiler soot blowing control method as described in any one of claims 1 to 7.

Images