Ash deposition monitoring and ash blowing optimization method and system for rotary air preheater

Abstract

The invention relates to the technical field of thermal power generation equipment state monitoring and intelligent operation and maintenance, in particular to an ash deposition monitoring and ash blowing optimization method and system for a rotary air preheater, and the method comprises the steps: S1, collecting a temperature data set and a visual image of the rotary air preheater; s2, constructing a dynamic temperature difference field, and determining an ash deposition area; s3, generating a dust deposition diagnosis result; and S4, soot blowing is carried out, the temperature data set and the visual image are collected again after soot blowing, and soot blowing operation is carried out again after soot blowing parameters are dynamically adjusted. The system comprises a temperature acquisition unit used for acquiring a temperature data set; the laser positioning subunit is used for calibrating the distance between the infrared array sensor and the rotor heat storage element in real time and dynamically correcting the temperature measurement value according to the distance change; an image acquisition unit; a data processing and control unit; an early warning unit; and a closed loop verification unit. The problems of high ash deposition misjudgment rate and large ash blowing energy consumption caused by single perception and one-sided model of a monitoring method in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of thermal power plant state monitoring and intelligent operation and maintenance, and in particular to a method and system for monitoring and optimizing ash deposition of a rotary air preheater. BACKGROUND

[0002] The rotary air preheater is a key equipment in the boiler system of a coal-fired power plant, and the problem of ash deposition thereof seriously affects the heat transfer efficiency and safe operation of the unit. At present, the monitoring of ash deposition of the rotary air preheater mainly adopts the mode of "single fixed infrared temperature measurement + temperature difference / clean factor", which can only obtain local instantaneous temperature and is difficult to generate a global dynamic temperature field with the rotation of the rotor (0.5-1.5 r / min), resulting in an ash deposition misjudgment rate of more than 35%. The blowing ash strategy generally adopts the "quantitative blowing of the whole area every 8 hours" mode, ignoring the uneven spatial distribution of ash deposition, causing waste of more than 30% of high-pressure steam, and accelerating the wear of the heat storage elements.

[0003] In a patent application with publication number CN118031881A and application date of February 2, 2024, a method for online monitoring of ash deposition thickness of a rotary air preheater is disclosed, which establishes a rotor micro-element heat transfer inversion model by finite difference method, and calculates the ash deposition thickness online using temperature gradient, realizing one-dimensional mapping of "temperature-thickness", but still relying on single temperature data, lacking visual verification, and being unable to output regional blowing ash strategy, with high misjudgment sensitivity. The genetic algorithm coupled with performance calculation software proposed by Hunan Electric Power in 2020 optimizes the blowing ash sequence, but it is difficult to be used online due to the lack of real-time image perception.

[0004] In a patent application with publication number CN116295044A and application date of February 15, 2023, an online monitoring method for flue ash thickness is disclosed, which uses laser or microwave ranging to directly obtain ash thickness, but is limited by high temperature, high dust and rotating structure, and cannot be used inside the rotary air preheater. Foreign literature focuses on numerical simulation or acoustic temperature measurement, and also does not form a practical system of "global temperature field + visual image" heterogeneous fusion.

[0005] In summary, the existing technologies generally have the defects of insufficient perception dimension, single model input, and passive blowing ash strategy, and there is no complete system disclosed for synchronous collection, fusion modeling and real-time self-optimization of blowing ash of "global temperature field-visual image". SUMMARY

[0006] The present application provides a method for monitoring and optimizing ash deposition of a rotary air preheater.

[0007] The application is implemented through the following technical scheme, and provides a method for monitoring and optimizing dust accumulation of a rotary air preheater, comprising the following steps: S1, arranging temperature data collection points around the rotary air preheater to collect a temperature data set, and arranging image collection points to collect visual images, wherein the collected temperature data set and visual images are synchronized with the rotation angle of the rotor of the rotary air preheater; S2, constructing a dynamic temperature difference field, calculating the radial temperature difference between the high-temperature side and the low-temperature side of the rotor at different rotation angles based on the temperature data set, generating a dynamic temperature difference field mapped with the rotation angle of the rotor, determining a dust accumulation area, analyzing the dynamic temperature difference field, identifying the area with a temperature difference higher than a preset threshold, and positioning the area to the position of the corresponding sector area of the rotor as the position of the dust accumulation area; S3, diagnosing the dust accumulation area through image recognition and fusion, calling the visual images associated with the positioned dust accumulation area, identifying the dust accumulation pattern through image analysis, and fusing the gradient information of the dynamic temperature difference field and the dust accumulation features of the images to generate a dust accumulation diagnosis result; and S4, controlling the soot blower to aim at the corresponding sector area of the rotor according to the position information in the dust accumulation diagnosis result, performing soot blowing, re-collecting the temperature data set and visual images after soot blowing, and performing the soot blowing operation again after dynamically adjusting the soot blowing parameters.

[0008] Specifically, the temperature data set in S1 comprises an upper data set and a lower data set, the upper data set is the temperature data of the high-temperature flue gas outlet end on the upper side of the rotary air preheater, and the lower data set is the temperature data of the low-temperature air inlet end on the lower side of the rotary air preheater.

[0009] Specifically, S2 is implemented according to the following steps:

[0010] S21, binding the temperature data set sampling time with the rotation angle θ of the rotor through the encoder signal of the rotor drive motor of the rotary air preheater, and establishing the mapping relationship between the angle θ and the temperature T;

[0011] S22, calculating the radial temperature difference of the upper temperature data and the lower temperature data at the same rotation angle θ, and generating a dynamic temperature difference field thermal map, wherein the radial temperature difference is calculated through the following formula,

[0012] ΔT(θ,i,j)=(T1,θ,i,j)-(T2,θ,i,j)

[0013] wherein ΔT is the radial temperature difference, θ is the rotation angle, i and j are array coordinates, i represents the radial direction coordinate from the center to the outer edge, j represents the circumferential coordinate of the rotation angle θ according to the rotation direction of the rotor, T1 is the upper temperature data, and T2 is the lower temperature data;

[0014] When ΔT significantly increases, the heat transfer of the dust accumulation area is weakened;

[0015] S23, based on temperature difference data, uses an improved Sobel operator to calculate the gradient magnitude;

[0016]

[0017] Where dx=dy is the distance between adjacent temperature measurement points, and G is the gradient magnitude. The larger the value of G, the more uneven the ash distribution and the more significant the thickness gradient.

[0018] The ash accumulation area is located by a region growing algorithm, where the seed point threshold of the region growing algorithm is G≥0.8℃ / mm. The region with a temperature difference ΔT higher than the threshold is identified and located in the corresponding sector area of ​​the rotor as the ash accumulation area.

[0019] Specifically, S3 is implemented according to the following steps:

[0020] S31, network structure optimization, based on U-Net, introduces a multi-scale feature fusion module and channel attention mechanism to segment image gray areas;

[0021] S32, using the obtained image training dataset;

[0022] S33, Ash accumulation feature extraction: Extract the area ratio A and average thickness H of the ash accumulation region from the segmentation results, and determine the calibration coefficient k;

[0023] Where A = dust accumulation area / total area, H = k × (255 - average gray level), k = 0.02 mm / gray level;

[0024] S34, classifying ash accumulation features, distinguishing ash accumulation types by image texture entropy E and the mean value of temperature feature ΔT;

[0025] When E≥6.5 and the mean ΔT≤30℃, it is considered fly ash accumulation;

[0026] When E≤5.0 and the mean ΔT value≥45℃, it is ammonium bisulfate ash accumulation;

[0027] S35, Generate the ash accumulation index model, construct a weighted fusion index S based on the above features, and obtain the degree of ash accumulation;

[0028]

[0029] The weighting coefficients are α=0.35, β=0.25, γ=0.3, and δ=0.1. Through optimization using a genetic algorithm, a quantitative assessment of the degree of ash accumulation is achieved, with the S value ranging from 0 to 100.

[0030] S36, a multi-dimensional early warning level classification, combines the ash accumulation index S, ash accumulation type and growth rate to divide the early warning into four levels;

[0031] The first level is the normal state, where S ≤ 20 and dS / dt ≤ 0.5 / h;

[0032] The second level is the state that needs attention, where 20 < S ≤ 40, or the fly ash fouling with dS / dt > 0.5 / h;

[0033] The third level is the moderate warning state, where 40 < S ≤ 70, or the ammonium bisulfate fouling with S > 20;

[0034] The fourth level is the emergency warning state, where S > 70, or dS / dt > 2 / h;

[0035] Where dS / dt is the growth rate, dt is the selected unit time, dS is the increase in the fouling index within this unit time, and h is the time unit of hour;

[0036] When S ≥ 37, multi-terminal linkage warning is carried out, and the fourth-level warning signal is synchronously pushed to the external warning device.

[0037] Specifically, the S4 is implemented in the following steps:

[0038] S41, regional precise positioning. Based on the rotation angle θ of the fouling core area, control the soot blower to be directed at the range of θ ± 5°;

[0039] S42, establish a calculation model for the soot blowing interval T according to the fouling growth rate and the unit load;

[0040]

[0041] Where k1 is the unit load condition coefficient, k2 is the fouling type condition coefficient, and P is the unit load;

[0042] S43, perform type matching and adaptation on the parameters, adjust the soot blowing parameters for different fouling types, and carry out soot blowing;

[0043] S44, conduct closed-loop verification and model iteration. After the soot blowing is completed, after a 20-minute interval, the rotor rotates 2 - 3 weeks, and the temperature and image data are collected again. Calculate the fouling index S' after soot blowing. If S' / S > 0.6, the cleaning effect is not good, then increase the pressure and extend the time to automatically correct the parameters, and update the strategy model to optimize k1 and k2 through reinforcement learning to ensure the convergence of the soot blowing effect.

[0044] Specifically, the training data set in S32 is as follows: 5000 sample images of fouling degree, cold-state illumination, and hot-state illumination with a thickness of 0.1 - 5 mm are collected, and expanded to 20000 images through data augmentation. The data expansion is rotation, scaling, and noise addition, so that the intersection over union of the model on the test set reaches 0.92.

[0045] Specifically, in S43, the soot blowing parameters are adjusted for different types of ash accumulation, as follows: for fly ash accumulation, the steam pressure is 1.2 MPa and the duration is 15 s; for ammonium bisulfate accumulation, the steam pressure is 0.9 MPa and the duration is 30 s.

[0046] A dust accumulation monitoring and soot blowing optimization system for a rotary air preheater, used to implement the above-mentioned method, includes: a temperature acquisition unit, which constitutes the temperature dataset acquisition point for acquiring temperature datasets; a laser positioning subunit, used to calibrate the distance between the infrared array sensor and the rotor heat storage element in real time, and dynamically correct the measured temperature value according to the distance change; an image acquisition unit, which is used to acquire visual images, and includes three industrial cameras, which are set on the observation hole on the air outlet side of the rotary air preheater; a data processing and control unit, which is configured to perform the construction of the dynamic temperature difference field, the location of the dust accumulation area, the diagnosis of image recognition and fusion, and targeted soot blowing; an early warning unit, which is used to receive the diagnostic results of the data processing and control unit and push alarm information to an external mobile terminal; and a closed-loop verification unit, which is used to restart the data acquisition and diagnostic process after the soot blowing operation is completed, and optimize the soot blowing operation according to the change in the degree of dust accumulation before and after soot blowing.

[0047] Specifically, the temperature detection unit includes 6 sets of infrared array sensors. 3 sets of infrared array sensors are set at the high-temperature flue gas outlet end on the upper side of the rotary air preheater and are equidistantly arranged along the circumference of the rotary air preheater shell. Each set contains 64 temperature measuring points. The other 3 sets of infrared array sensors are arranged at the low-temperature air inlet end on the lower side of the rotary air preheater. Their number and position correspond to the 3 sets of infrared array sensors set on the upper side of the rotary air preheater.

[0048] Specifically, the three industrial cameras include one wide-angle camera and two telephoto cameras.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) The method of the present invention collects temperature datasets and visual images, constructs dynamic temperature difference fields, determines the ash accumulation area, and diagnoses the ash accumulation area through image recognition and fusion, generating ash accumulation results to achieve a shift from qualitative description to quantitative diagnosis. In addition, the method also sets up closed-loop verification to iterate the model and achieve a leap from passive execution to active optimization. Through real-time verification of the effect after ash blowing, the operating parameters are automatically adjusted and the strategy model is iteratively optimized to achieve a self-optimization strategy of blowing when needed, blowing less for light ash, and blowing more for heavy ash, thereby reducing steam consumption and ash accumulation misjudgment rate, filling the gap in the domestic rotary air preheater "sensing-decision-execution" integrated closed-loop system, and greatly saving energy and reducing equipment wear while efficiently cleaning ash;

[0051] (2) Through the coordinated infrared array layout, the full-domain dynamic temperature field perception synchronized with the rotor rotation is realized, overcoming the limitations of single-point static temperature measurement. At the same time, combined with a vision system with dynamic tracking and environmental protection, the clear and stable capture of the ash accumulation pattern is achieved.

[0052] (3) This method integrates temperature and image heterogeneity, providing a multi-dimensional data foundation for diagnosis. By constructing a fusion analysis model of "temperature gradient-image features-ash type", the ash diagnosis is advanced from single parameter and qualitative judgment to a new stage of multi-source fusion and quantitative classification. The improved U-Net model enables the image segmentation accuracy to reach more than 0.92, and the classification diagnosis accuracy combined with temperature features exceeds 94%, fundamentally solving the problem of high misjudgment rate of traditional methods. Attached Figure Description

[0053] Figure 1 The thermal diagram of the temperature difference field of a rotary air preheater for monitoring ash accumulation and optimizing soot blowing according to the present invention.

[0054] Figure 2 This is a system structure diagram of a rotary air preheater ash accumulation monitoring and soot blowing optimization system according to the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature; in the description of this application, unless otherwise stated, "multiple" means two or more.

[0057] Please see Figure 1 and Figure 2 This invention provides a method for monitoring ash accumulation and optimizing soot blowing in a rotary air preheater, characterized by comprising:

[0058] S1, temperature data acquisition points are arranged around the rotary air preheater to collect temperature datasets, and image acquisition points are arranged to collect visual images. The collected temperature datasets and visual images are synchronized with the rotor rotation angle of the rotary air preheater. The upper dataset is the temperature data of the high-temperature flue gas outlet end on the upper side of the rotary air preheater, and the lower dataset is the temperature data of the low-temperature air inlet end on the lower side of the rotary air preheater.

[0059] S2, construct a dynamic temperature difference field. Based on the temperature dataset, calculate the radial temperature difference between the high-temperature side and the low-temperature side of the rotor at different rotation angles, generate a dynamic temperature difference field that maps to the rotor rotation angle, determine the dust accumulation area, analyze the dynamic temperature difference field, identify areas with temperature differences higher than a preset threshold, and locate them to the corresponding fan-shaped area of ​​the rotor as the dust accumulation area.

[0060] S2 is implemented according to the following steps.

[0061] S21, by using the encoder signal resolution of 1024 pulses / r of the rotor drive motor of the rotary air preheater, the sampling time of the temperature dataset is bound to the rotor rotation angle θ, and a mapping relationship between angle θ and temperature T is established. For example, when θ=30°, it corresponds to the second sector area of ​​the rotor.

[0062] S22 calculates the radial temperature difference between the upper and lower temperature data at the same rotation angle θ, and generates a dynamic temperature difference field thermogram as follows. Figure 1 The radial temperature difference is calculated using the following formula, which visually displays the localized low-temperature zone caused by dust accumulation. When heat transfer in the dust accumulation area weakens, ΔT increases significantly.

[0063] ΔT(θ,i,j)=(T1,θ,i,j)-(T2,θ,i,j)

[0064] Where ΔT is the radial temperature difference, θ is the rotation angle, i and j are array coordinates, i represents the radial coordinate from the center to the outer edge, j represents the circumferential coordinate according to the rotation angle θ of the rotor rotation direction, T1 is the upper temperature data, and T2 is the lower temperature data.

[0065] S23, based on temperature difference data, uses an improved Sobel operator to calculate the gradient magnitude;

[0066]

[0067] Where dx=dy is the distance between adjacent temperature measurement points, and G is the gradient magnitude. The larger the value of G, the more uneven the ash distribution and the more significant the thickness gradient. In this embodiment, dx=dy=50mm is the distance between adjacent temperature measurement points.

[0068] The ash accumulation area is located by a region growing algorithm. The seed point threshold of the region growing algorithm is G≥0.8℃ / mm. The region with a temperature difference ΔT higher than the threshold is identified and located in the corresponding sector area of ​​the rotor as the ash accumulation area. The region positioning accuracy is ≤±100mm.

[0069] S3, diagnosing the ash accumulation area through image recognition and fusion, calling up the visual image associated with the located ash accumulation area, identifying the ash accumulation morphology through image analysis, and fusing the gradient information of the dynamic temperature difference field with the ash accumulation features of the image to generate an ash accumulation diagnosis result; S3 is specifically implemented according to the following steps:

[0070] S31, network structure optimization, on the basis of traditional U-Net (image segmentation network), introduces a multi-scale feature fusion module to fuse 1 / 2, 1 / 4 and 1 / 8 downsampled feature maps, and a channel attention mechanism to highlight the gray-level difference features of the gray area to segment the gray area of ​​the image, so as to solve the segmentation blurring problem caused by the similarity between the gray area and the texture of the heat storage element;

[0071] S32 uses the obtained image training dataset; by collecting 5000 sample images with different dust accumulation levels, thickness range of 0.1-5mm, and different lighting conditions (cold or hot), and expanding to 20,000 images through data augmentation such as rotation, scaling, and noise addition, the model achieves a MIoU (Mean Intersection over Union) of 0.92 on the test set, while the traditional U-Net is only 0.78, thereby improving the accuracy of the model.

[0072] S33, Ash accumulation feature extraction: Extract the area ratio A and average thickness H of the ash accumulation region from the segmentation results, and determine the calibration coefficient k;

[0073] Where A = dust accumulation area / total area, and is obtained by inverting the gray value: H = k × (255 - mean gray value), from experimental data k = 0.02 mm / gray level;

[0074] S34, classifying ash accumulation features, distinguishing ash accumulation types by image texture entropy E and the mean value of temperature feature ΔT;

[0075] E is the image texture entropy, and its calculation formula is:

[0076]

[0077] Where:

[0078] P(i,j) represents the probability of pixel pairs with gray values i and j appearing in the image, which is calculated through the gray-level co-occurrence matrix.

[0079] i and j respectively represent different gray values.

[0080] This formula measures the randomness and complexity of the image gray-level distribution. The larger the entropy value, the more complex the image texture and the more information it contains.

[0081] When E≥6.5, it represents rough texture, and when the average value of ΔT≤30°C, it represents a relatively high thermal conductivity, which is fly ash fouling;

[0082] When E≤5.0, it represents smooth texture, and when the average value of ΔT≥45°C, it represents a low thermal conductivity and a significant heat insulation effect, which is ammonium bisulfate fouling;

[0083] S35. Generate a fouling index model, construct a weighted fusion index S based on the above features, and obtain the fouling degree;

[0084]

[0008] 3>

[0085] Among them, the weight coefficients are α = 0.35, β = 0.25, γ = 0.3, δ = 0.1, which are optimized through the genetic algorithm. Its objective function is to minimize the error with the manually measured fouling degree to achieve quantitative evaluation of the fouling degree. The value range of S is 0 - 100;

[0086] S36. Multidimensional early warning level division. Combine the fouling index S, fouling type, and growth rate to divide the early warning into four levels:

[0087] The first level is the normal state, S≤20, and the growth rate dS / dt≤0.5 / h, where h is the time unit in hours, dt is the selected unit time, and dS is the increase in the fouling index within this unit time;

[0088] The second level is the state that needs attention, 20 < S≤40, or fly ash fouling with dS / dt>0.5 / h;

[0089] The third level is the moderate early warning state, 40 < S≤70, or ammonium bisulfate fouling with S>20; Ammonium bisulfate fouling is easy to block and requires early intervention.

[0090] The fourth level is the emergency early warning state, S>70, or dS / dt>2 / h; The fouling grows rapidly and may cause fouling blockage.

[0091] S37, implement multi-terminal linkage early warning, and simultaneously push the Level 4 early warning signal to the external early warning device. In this embodiment, the external early warning device is the DCS system (audio-visual alarm) and the maintenance personnel's mobile APP (containing heat map + image of the dust accumulation area), and automatically retrieves the historical dust accumulation data of the area (trend of the past 72 hours) to assist maintenance decision-making.

[0092] S4, based on the location information in the ash accumulation diagnosis result, control the soot blower to aim at the corresponding rotor sector area for soot blowing. After soot blowing, re-acquire the temperature dataset and visual image, dynamically adjust the soot blowing parameters, and then perform the soot blowing operation again. S4 is specifically implemented according to the following steps:

[0093] S41, precise regional positioning, based on the rotation angle θ of the core dust accumulation area, assuming 8 soot blowers are arranged circumferentially, controlling the orientation of the soot blowers to align within a range of θ±5°, changing traditional full-area soot blowing to "targeted soot blowing", reducing the coverage area by 75%;

[0094] S42, Based on the ash accumulation growth rate and unit load, establish a calculation model for the soot blowing interval T;

[0095]

[0096] Where k1 is the unit load condition coefficient, k2 is the ash accumulation type condition coefficient, and P (MW) is the unit load;

[0097] For example, when the high load P=300MW, k1=0.8; for ammonium bisulfate ash, k2=1.5; and for fly ash, k2=1.0, to achieve dynamic adaptation of shortening the interval when the load is high and the growth is rapid.

[0098] S43, perform parameter type adaptation, adjust the soot blowing parameters for different types of ash accumulation, and perform soot blowing; for fly ash accumulation: steam pressure 1.2MPa impact force is preferred, duration 15s to avoid over-blowing; for ammonium bisulfate accumulation, steam pressure 0.9MPa to prevent component damage, duration 30s; ammonium bisulfate accumulation requires a longer softening and removal time.

[0099] S44. Perform closed-loop verification and model iteration. After the soot blowing is completed, wait 20 minutes, rotate the rotor 2-3 times, re-collect temperature and image data, and calculate the soot accumulation index S' after soot blowing. If S' / S>0.6, the cleaning effect is not good. Then increase the pressure and extend the time to automatically correct the parameters, such as increasing the pressure by 0.1MPa or extending the time by 5s. Update the strategy model and optimize k1 and k2 through reinforcement learning to ensure the soot blowing effect converges. The target is S' / S≤0.3.

[0100] The present invention also includes a dust accumulation monitoring and soot blowing optimization system for a rotary air preheater, comprising: a temperature acquisition unit, a laser positioning subunit, an image acquisition unit, a data processing and control unit, an early warning unit, and a closed-loop verification unit;

[0101] The temperature acquisition unit constitutes a temperature dataset acquisition point for collecting temperature datasets. The temperature detection unit includes 6 sets of infrared array sensors. 3 sets of infrared array sensors are set at the high-temperature flue gas outlet end on the upper side of the rotary air preheater and are equidistantly arranged along the circumference of the rotary air preheater shell. Each set contains 64 temperature measurement points. The other 3 sets of infrared array sensors are arranged at the low-temperature air inlet end on the lower side of the rotary air preheater. Their number and position correspond to the 3 sets of infrared array sensors set on the upper side of the rotary air preheater.

[0102] In this embodiment, three sets of infrared array sensors are equidistantly arranged along the circumference of the rotary air preheater shell at the high-temperature flue gas outlet end on the upper side of the rotary air preheater. The model is a customized 64-channel infrared thermopile array with a response wavelength of 8-14μm. Each set contains 16×4=64 temperature measurement points, covering a total of 64×3=192 points. The spacing between adjacent sensors is adapted according to the rotor diameter. For example, for a 300MW unit air preheater with a diameter of about 10m, the spacing is 1.5m to ensure coverage of the entire fan-shaped area of ​​the rotor, i.e., 0°-360°. At the low-temperature air inlet end on the lower side of the rotary air preheater, corresponding to the array position on the upper side, three sets of the same model of infrared array are arranged to form a synchronous monitoring with the upper side. The sampling frequency is ≥20Hz, and ≥800 sets of data can be collected in one rotation of the rotor. The temperature measurement range is -20℃~300℃, and the accuracy is ±0.3℃.

[0103] The laser positioning subunit is used to calibrate the distance between the infrared array sensor and the rotor heat storage element in real time, and dynamically correct the temperature measurement value according to the distance change.

[0104] In this embodiment, each infrared array integrates a laser positioning subunit with a wavelength of 650nm and a power ≤5mW. This subunit calibrates the distance error between the sensor and the rotor heat storage element in real time, ensuring the distance error is ≤±2mm. The measured temperature value is dynamically corrected based on distance changes, compensated for by the blackbody radiation law. ,in To calibrate the distance and resolve temperature measurement deviations caused by equipment vibration, traditional fixed installations can result in deviations of up to ±3℃.

[0105] The image acquisition unit is used to acquire visual images. The image acquisition unit includes three industrial cameras, which are installed on the observation hole on the air outlet side of the rotary air preheater.

[0106] In this embodiment, three high-definition industrial cameras with a resolution of 2560×1920 and a frame rate of ≥15fps are arranged at the observation port on the cold end air side of the air preheater, i.e., the air outlet side. A combination of one wide-angle panoramic camera and two telephoto close-up cameras is used.

[0107] The wide-angle camera, with an 8mm focal length, covers the entire 360° area of ​​the rotor, capturing the macroscopic distribution of accumulated dust in real time. The telephoto camera, with a 50mm focal length, uses a servo motor-driven rotating gimbal to dynamically track the "high temperature difference area," or potential dust accumulation area, marked by an infrared array, enabling detailed capture of dust morphology in localized areas with a resolution ≤0.1mm / pixel. The camera lens integrates a dual-airflow purging + electrically heated anti-fog film: the outer layer uses 0.4MPa compressed air to purge dust from the lens surface, while the inner layer uses nitrogen with a dew point ≤-40℃ to isolate moisture. Combined with a 20W silicone rubber heating element, the temperature is controlled at 50±5℃, ensuring an image clarity retention rate ≥95% in environments with 85% humidity and a dust concentration of 30g / Nm³, compared to less than 50% clarity degradation in traditional unprotected designs.

[0108] The data processing and control unit is configured to perform the construction of the dynamic temperature difference field, the location of the ash accumulation area, the diagnosis of image recognition and fusion, and targeted soot blowing; the early warning unit is used to receive the diagnostic results of the data control unit and push visual alarm information containing thermal maps and images of the ash accumulation area to the distributed control system and mobile terminal; the closed-loop verification unit is used to restart the data acquisition and diagnostic process at a predetermined time after the soot blowing operation is completed, and adaptively optimize the soot blowing parameters according to the change in the degree of ash accumulation before and after soot blowing.

[0109] Implementation Case:

[0110] This embodiment selects a 600MW unit rotary air preheater and performs the following operations:

[0111] S1: Temperature data acquisition points and image acquisition points are set up. Three sets of infrared array sensors with a total of 192 temperature measurement points are installed on the upper side of the rotary air preheater, and three sets of infrared array sensors are installed on the lower side of the rotary air preheater. Three cameras are arranged on the cold end air side. The data processing server GPU is such as NVIDIA A100. The system architecture diagram is as follows. Figure 2 As shown;

[0112] S2: Determine the dust accumulation area;

[0113] S3: Diagnosing ash accumulation areas. In practice, the accuracy rate for identifying 3mm thick ammonium bisulfate ash accumulation areas reached 94%, compared to only 68% using the traditional temperature difference method, with a positioning deviation ≤80mm. The accuracy rate of the ash accumulation diagnosis was obtained through the following methods: 1) Establishing a measured benchmark: After shutdown, the actual thickness / mass of the ash accumulation was obtained through manual cleaning and weighing, endoscopic observation, and ultrasonic thickness measurement, serving as a benchmark for comparison of the diagnostic results. 2) Algorithm output compared with the benchmark: The ash accumulation status was output using the valve algorithm and matched region by region with the measured benchmark to calculate the accuracy rate (number of correct diagnoses / total number of diagnoses).

[0114] S4: Perform soot blowing. After soot blowing, dynamically adjust the soot blowing parameters and perform soot blowing again. The soot blowing efficiency was optimized during implementation. The results showed that the soot blowing steam consumption was reduced by 32% (the soot blowing steam consumption was obtained by reading the instantaneous flow rate during soot blowing at the DCS terminal in real time, combined with the duration and frequency of a single soot blowing, to accumulate the steam consumption per unit time (e.g., daily / monthly)). It decreased from 8t / h to 5.4t / h. The average resistance of the rotary air preheater decreased by 180Pa, the specific heat transfer efficiency increased by 4.2%, and no unplanned shutdowns due to ash accumulation occurred.

[0115] By implementing and coordinating the layout and dynamic calibration of upper and lower infrared array sensors, and deploying cameras on the cold-end air side for multi-view tracking and environmental adaptation protection, the synchronous and clear perception of the rotor's full-domain temperature field and ash accumulation morphology is achieved, with 100% spatial coverage and a 45% improvement in environmental adaptability. In addition, the synchronization of component rotation angle and temperature in time and space, the improved U-Net and channel attention to form an ash accumulation classification diagnostic model, achieve an ash accumulation segmentation mIoU of 0.92 and a classification accuracy of 94%, solving the problem of misjudgment by traditional single data. Finally, three-dimensional optimization of region, time sequence and parameters, closed-loop verification and model iteration are carried out, resulting in a 32% reduction in soot blowing steam consumption and an ash removal effect converged to S' / S≤0.3, achieving precise, energy-saving and adaptive soot blowing control, greatly saving energy and reducing equipment wear.

[0116] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. This application is not limited to the exact structures described above and illustrated in the accompanying drawings, and it should not be considered that the specific implementation of this application is limited to these descriptions. For those skilled in the art, various changes and modifications made without departing from the concept of this application should be considered to fall within the protection scope of this application.

Claims

1. A method of soot monitoring and soot blowing optimization for a rotary air preheater, characterized in that, The method comprises the following steps: S1, arranging temperature data collection points around the rotary air preheater to collect a temperature data set, and arranging image collection points to collect visual images, the collected temperature data set and visual images being synchronized with the rotation angle of the rotor of the rotary air preheater; S2, constructing a dynamic temperature difference field, calculating the radial temperature difference between the high-temperature side and the low-temperature side of the rotor at different rotation angles based on the temperature data set, generating a dynamic temperature difference field mapped with the rotation angle of the rotor, determining the ash deposition area, analyzing the dynamic temperature difference field, identifying the area with a temperature difference higher than a preset threshold, and positioning the area to the corresponding sector area of the rotor as the position of the ash deposition area; S3, diagnosing the ash deposition area through image recognition and fusion, calling the visual image associated with the positioned ash deposition area, identifying the ash deposition morphology through image analysis, and fusing the gradient information of the dynamic temperature difference field and the ash deposition characteristics of the image to generate an ash deposition diagnosis result; S4, according to the position information in the ash deposition diagnosis result, controlling the soot blower to aim at the corresponding sector area of the rotor for soot blowing, re-collecting the temperature data set and visual images after soot blowing, and performing the soot blowing operation again after dynamically adjusting the soot blowing parameters.

2. The method of claim 1, wherein: The temperature data set in S1 comprises an upper data set and a lower data set, the upper data set being the temperature data of the high-temperature flue gas outlet end on the upper side of the rotary air preheater, and the lower data set being the temperature data of the low-temperature air inlet end on the lower side of the rotary air preheater.

3. The ash deposition monitoring and soot blowing optimization method of the rotary air preheater according to claim 2, characterized in that S2 is implemented according to the following steps: S21, binding the temperature data set sampling time with the rotation angle θ of the rotor through the encoder signal of the rotor drive motor of the rotary air preheater, and establishing a mapping relationship between the angle θ and the temperature T; S22, calculating the radial temperature difference of the upper temperature data and the lower temperature data at the same rotation angle θ, and generating a dynamic temperature difference field thermal map, the radial temperature difference being calculated by the following formula, ΔT(θ, i, j) = (T1, θ, i, j) - (T2, θ, i, j) where ΔT is the radial temperature difference, θ is the rotation angle, i and j are array coordinates, i represents the radial direction coordinate from the center to the outer edge, j represents the circumferential coordinate of the rotation angle θ according to the rotation direction of the rotor, T1 is the upper temperature data, and T2 is the lower temperature data; when ΔT significantly increases, the heat transfer of the ash deposition area is weakened; S23, calculating the gradient modulus value based on the temperature difference data using the improved Sobel operator; where dx = dy is the distance between adjacent temperature measurement points, and G is the gradient modulus value, a larger G value indicates that the ash deposition distribution is more uneven and the thickness gradient is more significant; locating the ash deposition area through the region growing algorithm, wherein the seed point threshold of the region growing algorithm is G≥0.8℃ / mm, and the area with a temperature difference ΔT higher than the threshold is positioned to the corresponding sector area of the rotor as the ash deposition area.

4. The ash deposition monitoring and soot blowing optimization method of the rotary air preheater according to claim 3, characterized in that S3 is implemented according to the following steps: S31, network structure optimization, on the basis of U-Net, introduce multi-scale feature fusion module and channel attention mechanism to segment the image dust accumulation area; S32, through the obtained image training data set; S33, dust accumulation feature extraction, extract the area ratio A and the average thickness H of the dust accumulation area from the segmentation result, and determine the calibration coefficient k; Wherein A=dust accumulation area / total area, H=k x (255-gray mean), k=0.02mm / gray level; S34, dust accumulation feature classification, distinguish dust accumulation type through image texture entropy E and temperature feature ΔT mean; When E≥6.5, ΔT mean≤30℃, it is fly ash dust; When E≤5.0, ΔT mean≥45℃, it is ammonium bisulfate dust; S35, generate dust accumulation index model, construct weighted fusion index S based on the above features, and obtain the dust accumulation degree; Wherein the weight coefficient α=0.35, β=0.25, γ=0.3, δ=0.1, and the quantitative evaluation of dust accumulation degree is realized through genetic algorithm optimization, wherein the S value range is 0-100; S36, multi-dimensional warning level division, combining dust accumulation index S, dust accumulation type and growth rate, the warning is divided into four levels; The first level is normal state, S≤20, dS / dt≤0.5 / h; The second level is the state that needs attention, 20<S≤40, or dS / dt>0.5 / h fly ash dust; The third level is the moderate warning state, 40<S≤70, or S>20 ammonium bisulfate dust; The fourth level is the emergency warning state, S>70, or dS / dt>2 / h; Wherein dS / dt is the growth rate, dt is the selected unit time, dS is the growth amount of dust accumulation index in the unit time, and h is the time unit hour; S37, multi-terminal linkage warning, the fourth level warning signal is pushed to the external warning device synchronously.

5. The dust accumulation monitoring and soot blowing optimization method of the rotary air preheater according to claim 4, characterized in that, The S4 is specifically implemented according to the following steps: S41, accurate positioning of the area, based on the rotation angle θ of the dust accumulation core area, the soot blower is controlled to be directional and aligned within the range of θ±5°; S42, according to the dust accumulation growth rate and the unit load, a calculation model of the soot blowing interval T is established; Wherein k1 is the unit load working condition coefficient, k2 is the dust accumulation type working condition coefficient, and P is the unit load; S43, parameter classification and adaptation, adjust the soot blowing parameters for different dust accumulation types, and perform soot blowing; S44, closed loop verification and model iteration, 20 minutes after soot blowing, the rotor rotates 2-3 weeks, the temperature and image data are reacquired, the dust accumulation index S' after soot blowing is calculated, if S' / S>0.6, the cleaning effect is poor, then the pressure is increased and the time is prolonged to automatically correct the parameters, and the strategy model is updated, and k1 and k2 are optimized through reinforcement learning, so as to ensure the convergence of the soot blowing effect.

6. The method of claim 4, wherein, In S32, the training data set is trained, and the specific process is as follows: 5000 sample images under the conditions of dust accumulation degree of 0.1-5mm, cold light and hot light are collected, and the number of images is expanded to 20000 through data enhancement, and the data expansion is rotation, scaling and noise addition, so that the intersection over union of the model on the test set reaches 0.

92.

7. The method of claim 5, wherein the method further comprises: In S43, the soot blowing parameters are adjusted for different types of dust accumulation, and the specific process is as follows: for fly ash dust accumulation, the steam pressure is 1.2MPa, and the duration is 15s; for ammonium bisulfate dust accumulation, the steam pressure is 0.9MPa, and the duration is 30s.

8. A system for monitoring and optimizing the soot blowing of a rotary air preheater, for implementing the method for monitoring and optimizing the soot blowing of a rotary air preheater according to any one of claims 1 to 7, characterized in that, It comprises: A temperature acquisition unit, which constitutes a temperature data set acquisition point for acquiring a temperature data set, A laser positioning subunit for real-time calibration of the distance between the infrared array sensor and the rotor heat storage element, and dynamic correction of the temperature measurement value according to the distance change; An image acquisition unit, which is used for acquiring visual images, and comprises three industrial cameras arranged on the observation holes on the air outlet side of the rotary air preheater; A data processing and control unit configured to perform the construction of the dynamic temperature difference field, the positioning of the dust accumulation area, the diagnosis of image recognition and fusion, and targeted soot blowing; An early warning unit for receiving the diagnosis results of the data control unit and pushing alarm information to external mobile terminals; A closed-loop verification unit for restarting the data acquisition and diagnosis process after the soot blowing operation is completed, and optimizing the soot blowing operation according to the change amount of the dust accumulation degree before and after soot blowing.

9. The soot monitoring and soot-blowing optimization system for a rotary air preheater of claim 8, wherein: The temperature detection unit comprises six groups of infrared array sensors, three groups of which are arranged at the high-temperature flue gas outlet end of the rotary air preheater on the upper side, arranged equidistantly along the circumference of the rotary air preheater shell, and each group contains 64 temperature measurement points. The other three groups of infrared array sensors are arranged at the low-temperature air inlet end of the rotary air preheater on the lower side, and the number and position correspond to the three groups of infrared array sensors arranged on the upper side of the rotary air preheater.

10. The soot monitoring and soot blowing optimization system of the rotary air preheater according to claim 8, characterized in that: The three industrial cameras comprise one wide-angle camera and two long-focus cameras.

Images