Method and system for optimizing soot blowing control for anti-coking of a boiler of a thermal power unit

Abstract

The application provides a method and system for preventing coking of a boiler of a thermal power unit, and relates to the technical field of thermal power generation, and comprises the following steps: collecting heating surface temperature and flue gas flow field data, identifying a coking area, constructing a coking risk assessment matrix, determining a soot blower starting sequence and a time interval distribution scheme, and combining steam pipe network pressure data to generate a soot blowing control instruction sequence. The application can accurately identify the coking position, optimize the soot blowing sequence and intensity, improve the soot blowing efficiency, prolong the service life of the equipment, and reduce energy consumption.

Description

Technical Field

[0001] This invention relates to the field of thermal power generation technology, and in particular to an optimized method and system for soot blowing control to prevent coking in boilers of thermal power units. Background Technology

[0002] During long-term operation, ash, carbon powder, and fusible substances in the fuel accumulate on the heating surfaces of thermal power boilers, forming coke. This leads to decreased heat transfer efficiency, localized high temperatures, and even safety accidents. To maintain the safe and stable operation of the boiler, soot blowers are needed to periodically remove the coke. Traditional soot blowing control relies mainly on manual experience and fixed cycles, which cannot respond promptly to actual coking conditions and is difficult to optimize the utilization efficiency of soot blowing resources. As thermal power units develop towards larger sizes and higher parameters, boiler structures are becoming increasingly complex, heating areas are increasing, and the coking problem is becoming more severe, placing higher demands on soot blowing control.

[0003] Existing technologies still suffer from a lack of precise identification and quantitative assessment of coking conditions. This makes it difficult to accurately identify the specific location, extent, and severity of coking, resulting in the inability to accurately deploy soot blowing resources to the areas where they are truly needed. Furthermore, these technologies fail to fully consider multi-dimensional information such as boiler combustion conditions, heating surface temperature distribution, and flue gas flow characteristics. They also lack a dynamic model for coking formation and propagation, making it impossible to achieve proactive prevention and precise intervention. Additionally, they lack systematic optimization, failing to make overall plans based on the severity of coking, propagation risks, and system constraints. Consequently, they struggle to achieve optimal cleaning results within limited steam resources and time windows while ensuring the stability of steam network pressure. Summary of the Invention

[0004] This invention provides a soot blowing control optimization method and system for preventing coking in thermal power unit boilers, which can at least solve some of the problems existing in the prior art.

[0005] A first aspect of this invention provides a soot blowing control optimization method for preventing coking in thermal power unit boilers, comprising: The boiler's heating surface temperature data and flue gas flow field pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve. The pressure deviation vector is determined by combining the pre-acquired measured pressure distribution data. The coking degree quantification index is obtained by combining the suspected coking areas. Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. The heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the coking degree quantification index. The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed by combining it with the pre-acquired flue gas flow path. The coking propagation coefficient is calculated based on the directed weighted graph and the coking acceleration node is determined by the Markov chain. The soot blower start sequence is determined based on the coking acceleration node and a time interval allocation scheme is generated by combining it with the coking propagation coefficient. Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Calculate the duration of action based on the upper limit of available steam flow and the sootblower start-up sequence. Determine the total execution time based on the duration of action and the time interval allocation scheme, and determine whether it exceeds the time window constraint. If it does, compress the duration proportionally to obtain a corrected duration and combine it with the sootblower start-up sequence to generate a sootblowing control command sequence for execution.

[0006] In one alternative implementation, The boiler's heating surface temperature data and flue gas flow pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow pressure distribution data to reconstruct the pressure baseline curve, including: The distributed temperature sensor array collects multi-point temperature data of the boiler heating surface at a preset sampling frequency, and the pressure sensor array collects multi-point pressure data of the flue gas flow field at a preset sampling frequency. Based on the timestamps and spatial coordinates corresponding to the multi-point temperature data, a spatiotemporal temperature matrix is ​​constructed. The temperature difference between each measuring point in the spatiotemporal temperature matrix and its adjacent measuring points is calculated. The temperature difference is used as a distance metric to determine the density reachability of the measuring points. The density-reachable measuring points are merged into multiple temperature clusters. The centroid temperature and spatial range of each temperature cluster are calculated. It is determined whether the centroid temperature is higher than a preset temperature threshold and whether the spatial range is greater than a preset range threshold. If both conditions are met, the heated surface area corresponding to the current temperature cluster is marked as a suspected coking area. The multi-point pressure data is converted into a frequency domain signal and subjected to Fourier transform to obtain a spectral distribution. The low-frequency components of the spectral distribution are extracted and the main frequency components are retained. The main frequency components are then subjected to inverse Fourier transform to reconstruct the pressure reference curve.

[0007] In one alternative implementation, By combining pre-acquired measured pressure distribution data to determine the pressure deviation vector, and by combining the suspected coking area to obtain quantitative indicators of coking degree, including: Obtain the measured pressure distribution data at the time corresponding to the pressure reference curve, align the pressure reference curve and the measured pressure distribution data according to spatial coordinates, calculate the difference between the pressure value of the pressure reference curve at each spatial coordinate point and the pressure value of the measured pressure distribution data at the corresponding spatial coordinate point, and arrange them according to the spatial coordinate order to obtain the pressure deviation vector. Based on the spatial coordinates of the measured pressure distribution data, a spatial grid of the heated surface is constructed. The pressure deviation vector is mapped onto the spatial grid of the heated surface according to the spatial coordinates to form a pressure deviation distribution field. The spatial coordinate range of the suspected coking area is extracted. The pressure deviation values ​​of the grid nodes corresponding to the spatial coordinate range are extracted from the pressure deviation distribution field, and spatial gradient calculation is performed to obtain the pressure deviation gradient vector. Principal component analysis is performed on the pressure deviation gradient vector to obtain the dominant gradient direction and the variance contribution rate corresponding to each principal component. The angle between the dominant gradient direction and the flue gas flow direction is calculated to obtain the flow direction deviation. The variance contribution rate is used as a weight to perform a weighted summation of the pressure deviation values ​​of each grid node in the suspected coking area. The coking degree quantification index is obtained by combining the flow direction deviation and the spatial range of the suspected coking area.

[0008] In one alternative implementation, Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. Heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the aforementioned coking degree quantification index. This includes: Historical boiler combustion conditions are obtained from the historical database, and the injection angle parameters and injection speed parameters of each burner are extracted from the historical boiler combustion conditions. Taking the spatial position of the burner as the starting point of the ray, the injection angle parameter determines the propagation direction of the ray, the injection speed parameter determines the propagation step size of the ray, and the ray is gradually tracked along the propagation direction and the intersection position of the ray with each spatial coordinate point of the heated surface is recorded. The cumulative number of times each spatial coordinate point is passed by the ray is counted and the heat load intensity value is determined. The heat load intensity values ​​of all spatial coordinate points are arranged according to the spatial coordinates to obtain the heat load intensity distribution. The spatial coordinate range of the suspected coking area is obtained. The heat load intensity value corresponding to the spatial coordinate range is extracted from the heat load intensity distribution and matched with the coking degree quantification index according to the spatial coordinates to obtain a paired dataset. The heat load intensity value in the paired dataset is divided into intervals to obtain the heat load intensity level. The coking degree quantification index in the paired dataset is divided into intervals to obtain the coking degree level. The heat load intensity level is used as the row index and the coking degree level is used as the column index. The coking risk assessment matrix is ​​constructed by traversing each spatial coordinate point in the paired dataset.

[0009] In one alternative implementation, The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed using pre-acquired flue gas flow paths. Based on the directed weighted graph, the coking propagation coefficient is calculated, and coking acceleration nodes are determined using Markov chains. Based on the coking acceleration nodes, a sootblower activation sequence is determined, and a time interval allocation scheme is generated using the coking propagation coefficient, including: The elements in the coking risk assessment matrix are sorted in descending order according to their numerical values ​​to obtain an element sequence. The dependencies between different matrix elements are determined based on the heat load intensity level and coking degree level corresponding to each element, and the element sequence is topologically sorted to obtain a priority sequence. Project the spatial coordinate points corresponding to the elements in the priority sequence along the flue gas flow path to obtain the projection position. Use the spatial coordinate points as graph nodes, determine the adjacent relationship between the graph nodes according to the projection position as the edge connection relationship, and construct a directed weighted graph using the elements of each spatial coordinate point in the coking risk assessment matrix as the node weight value. The outgoing edge propagation intensity is obtained by traversing the directed weighted graph and calculating the ratio of the weight value corresponding to the current graph node to the weight value corresponding to the downstream node. The outgoing edge propagation intensity of each graph node is summed to obtain the coking propagation coefficient. A Markov chain is constructed with the graph nodes as states and the coking propagation coefficient as the state transition probability, and a steady-state probability distribution is obtained by iterative calculation. Based on the steady-state probability distribution, coking acceleration nodes are determined. The spatial coordinates of each coking acceleration node are extracted and the corresponding sootblower number is queried. The sootblower start sequence is determined according to the order of the coking acceleration nodes on the flue gas flow path and the sootblower number. The time interval is determined by combining the coking propagation coefficient between adjacent coking acceleration nodes and a time interval allocation scheme is generated.

[0010] In one alternative implementation, Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the sootblower start-up sequence, calculate the duration of operation, including: The steam pipeline network monitoring system acquires steam pipeline network pressure fluctuation data within a preset time period and performs time-domain decomposition to obtain a pressure reference component and a pressure fluctuation component. Spectral analysis is performed on the pressure fluctuation component to obtain the dominant frequency component and its corresponding amplitude. Based on the amplitude corresponding to the dominant frequency component, the pressure fluctuation intensity is calculated using an energy accumulation method. The steady-state available steam flow rate is calculated based on the pressure reference component. The flow loss corresponding to the safety margin is calculated based on the pressure fluctuation intensity to obtain the fluctuation margin. Based on the steady-state available steam flow rate and the fluctuation margin, the upper limit of the available steam flow rate is obtained using a margin deduction method. Extract the number of each sootblower in the sootblower start-up sequence and query the corresponding rated steam consumption and rated operating time. Obtain the coking propagation coefficient of the coking acceleration node corresponding to each sootblower and perform normalization processing to obtain the coking propagation weight corresponding to each sootblower. Based on the rated steam consumption and the coking propagation weight, obtain the weighted steam consumption by weighted summation. Determine the sufficiency of the available steam flow rate by comparing the upper limit of available steam flow rate with the weighted steam consumption to obtain the determination result. Based on the determination result, combine the rated operating time and the coking propagation weight to solve for the operating duration of each sootblower.

[0011] In one alternative implementation, Based on the duration of the action and the time interval allocation scheme, the total execution time is determined and it is judged whether it exceeds the time window constraint. If it does, the duration is compressed proportionally to obtain a corrected duration, and combined with the soot blower start sequence, a soot blowing control command sequence is generated and issued for execution, including: The time interval between adjacent sootblowers in the sootblower start-up sequence is extracted from the time interval allocation scheme and accumulated with the duration of action to obtain the total execution time. The current boiler operating condition data is obtained and the current load status and flue gas temperature are extracted. The corresponding time window constraint upper limit is queried according to the current load status and the flue gas temperature. It is determined whether the total execution time is greater than the time window constraint upper limit. If it is greater, the ratio coefficient between the total execution time and the time window constraint upper limit is calculated. The compression operation is performed based on the duration of action and the ratio coefficient to obtain the corrected duration. The steady-state probability value of the coking acceleration node corresponding to each sootblower is extracted and the priority weight is calculated. Based on the priority weight, the correction duration is redistributed and adjusted to obtain the optimal duration. Based on the optimal duration and the sootblower number in the sootblower start sequence, the start time and stop time of each sootblower are generated sequentially and encapsulated to obtain sootblowing control instructions. The sootblowing control instructions are arranged in the order of start time to generate a sootblowing control instruction sequence and sent to the sootblower control unit for execution through the industrial control network.

[0012] A second aspect of the present invention provides a soot blowing control optimization system for preventing coking in thermal power unit boilers, comprising: The coking sensing module is used to collect boiler heating surface temperature data and flue gas flow field pressure distribution data. It performs spatiotemporal clustering analysis on the heating surface temperature data to identify suspected coking areas, performs Fourier transform on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve, determines the pressure deviation vector by combining the pre-acquired measured pressure distribution data, and obtains the coking degree quantification index by combining the suspected coking areas. The risk assessment module is used to obtain historical boiler combustion conditions and extract burner injection angle parameters, calculate heat load intensity distribution through ray tracing algorithm, and construct a coking risk assessment matrix by combining the coking degree quantification index. The sequence generation module is used to perform topological sorting on the coking risk assessment matrix and construct a directed weighted graph in combination with the pre-acquired flue gas flow path. Based on the directed weighted graph, the coking propagation coefficient is calculated and the coking acceleration node is determined through a Markov chain. Based on the coking acceleration node, the soot blower start sequence is determined and a time interval allocation scheme is generated in combination with the coking propagation coefficient. The soot blowing control module is used to acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the soot blower start sequence, it calculates the duration of action. Based on the duration of action and the time interval allocation scheme, it determines the total execution time and determines whether it exceeds the time window constraint. If it exceeds the constraint, it compresses the duration proportionally to obtain a corrected duration and combines it with the soot blower start sequence to generate a soot blowing control command sequence for execution.

[0013] A third aspect of the present invention provides an electronic device, comprising: A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] In this invention, by performing spatiotemporal clustering analysis on the temperature data of the heated surface and Fourier transform reconstruction on the flue gas flow field pressure distribution data, the coking area is accurately identified and quantified. Based on the ray tracing algorithm, the heat load intensity distribution is calculated and a coking risk assessment matrix is ​​constructed in combination with the coking degree quantification index, realizing a comprehensive assessment of boiler coking risk and improving the accuracy of coking early warning. By performing topological sorting on the coking risk assessment matrix and constructing a directed weighted graph, and combining it with Markov chains to determine coking acceleration nodes, the problem of identifying coking propagation paths is solved, making the sootblower start-up sequence more targeted. Based on the steam pipeline pressure fluctuation data, the upper limit of available steam flow is calculated, and time window constraint judgment is performed in combination with the time interval allocation scheme, realizing the optimization of sootblowing control under limited resource conditions, which not only ensures the sootblowing effect but also avoids the waste of steam resources. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the optimized soot blowing control method for preventing coking in thermal power unit boilers according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the coking risk assessment and sootblower scheduling optimization process of the sootblowing control optimization method for preventing coking in thermal power unit boilers, as described in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0019] Figure 1 This is a flowchart illustrating the optimized soot blowing control method for preventing coking in thermal power unit boilers according to an embodiment of the present invention. Figure 1 As shown, the method includes: The boiler's heating surface temperature data and flue gas flow field pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve. The pressure deviation vector is determined by combining the pre-acquired measured pressure distribution data. The coking degree quantification index is obtained by combining the suspected coking areas. Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. The heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the coking degree quantification index. The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed by combining it with the pre-acquired flue gas flow path. The coking propagation coefficient is calculated based on the directed weighted graph and the coking acceleration node is determined by the Markov chain. The soot blower start sequence is determined based on the coking acceleration node and a time interval allocation scheme is generated by combining it with the coking propagation coefficient. Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Calculate the duration of action based on the upper limit of available steam flow and the sootblower start-up sequence. Determine the total execution time based on the duration of action and the time interval allocation scheme, and determine whether it exceeds the time window constraint. If it does, compress the duration proportionally to obtain a corrected duration and combine it with the sootblower start-up sequence to generate a sootblowing control command sequence for execution.

[0020] In one alternative implementation, The boiler's heating surface temperature data and flue gas flow pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow pressure distribution data to reconstruct the pressure baseline curve, including: The distributed temperature sensor array collects multi-point temperature data of the boiler heating surface at a preset sampling frequency, and the pressure sensor array collects multi-point pressure data of the flue gas flow field at a preset sampling frequency. Based on the timestamps and spatial coordinates corresponding to the multi-point temperature data, a spatiotemporal temperature matrix is ​​constructed. The temperature difference between each measuring point in the spatiotemporal temperature matrix and its adjacent measuring points is calculated. The temperature difference is used as a distance metric to determine the density reachability of the measuring points. The density-reachable measuring points are merged into multiple temperature clusters. The centroid temperature and spatial range of each temperature cluster are calculated. It is determined whether the centroid temperature is higher than a preset temperature threshold and whether the spatial range is greater than a preset range threshold. If both conditions are met, the heated surface area corresponding to the current temperature cluster is marked as a suspected coking area. The multi-point pressure data is converted into a frequency domain signal and subjected to Fourier transform to obtain a spectral distribution. The low-frequency components of the spectral distribution are extracted and the main frequency components are retained. The main frequency components are then subjected to inverse Fourier transform to reconstruct the pressure reference curve.

[0021] A distributed temperature sensor array is deployed on the boiler's heating surface, with a sensor spacing of 0.5 to 1 meter, forming a grid-like coverage structure. Pressure sensor arrays are installed at different heights and circumferential positions in the flue gas passage. The sampling frequency of the temperature sensors is set to 1 Hz, and the sampling frequency of the pressure sensors is set to 10 Hz to ensure the timeliness and accuracy of the data.

[0022] The temperature sensor array collects data including temperature values ​​and their corresponding timestamps and spatial coordinates. Taking the superheater area as an example, a 10×8 sensor grid is arranged, with a total of 80 temperature measurement points. Each measurement point's data includes the temperature value, the acquisition time, and three-dimensional spatial coordinates. Data from 30 consecutive minutes is organized according to timestamps and spatial locations to construct a spatiotemporal temperature matrix. The spatiotemporal temperature matrix has dimensions of 1800×80, where 1800 represents the number of time points acquired at a frequency of 1 Hz within 30 minutes, and 80 represents the number of spatial measurement points.

[0023] For each measuring point in the spatiotemporal temperature matrix, the temperature difference between the current measuring point and its adjacent measuring points is calculated. The temperature difference is calculated using the absolute temperature difference between the measuring points. For example, if the temperature at a measuring point is 520 degrees Celsius and the temperature at an adjacent measuring point is 490 degrees Celsius, the temperature difference is 30 degrees Celsius. A temperature difference threshold of 25 degrees Celsius and a spatial distance threshold of 1.5 meters are set. If the temperature difference between two measuring points is less than the temperature difference threshold and the spatial distance is less than the spatial distance threshold, then the density of these two measuring points is considered reachable.

[0024] Based on density reachability, all measuring points are clustered and merged. Starting with the measuring point with the highest temperature, its density reachability with other measuring points is determined sequentially, and all density-reachable measuring points are grouped into a single temperature cluster. This process is repeated until all measuring points are grouped, ultimately forming multiple temperature clusters. In a practical application, a certain detection resulted in four temperature clusters, containing 15, 12, 9, and 7 measuring points respectively.

[0025] For each temperature cluster, the centroid temperature and spatial range are calculated. The centroid temperature is the average temperature of all measuring points within the cluster. The spatial range is the distance between the two farthest measuring points within the cluster. For example, the average temperature of 15 measuring points in a certain temperature cluster is 540 degrees Celsius, and the distance between the two farthest points is 3.2 meters. The temperature threshold for determining coking is set to 530 degrees Celsius, and the range threshold is set to 2.5 meters. If the centroid temperature of a temperature cluster is higher than the temperature threshold and the spatial range is greater than the range threshold, then the area is marked as a suspected coking area. In this embodiment, this temperature cluster is determined to be a suspected coking area.

[0026] For multi-point pressure data collected by the pressure sensor array, the raw data is preprocessed, including outlier removal and signal smoothing. Taking pressure sensor data from a certain flue section as an example, the sampling frequency is 10 Hz, and continuous acquisition for 10 minutes yields pressure data at 6000 time points. The aforementioned pressure data is then converted into a frequency domain signal for spectral analysis.

[0027] Frequency domain transformation is achieved through Fourier transform, mapping time series data to frequency space. For 6000 sampling points, the transformation yields a spectral distribution in the range of 0 to 5 Hz. From this spectral distribution, low-frequency components in the 0 to 0.5 Hz range are extracted, and frequency components with amplitudes exceeding 20% ​​of the total energy are retained, typically two to three dominant frequency components. For example, in one analysis, the three dominant frequency components of 0.05 Hz, 0.12 Hz, and 0.32 Hz were retained, accounting for 75% of the total energy.

[0028] The extracted main frequency components were subjected to inverse Fourier transform to reconstruct the pressure baseline curve. The pressure baseline curve reflects the main pressure fluctuation characteristics of the flue gas flow field, removing high-frequency noise and random disturbances. The reconstructed pressure baseline curve was used for subsequent analysis of the airflow characteristics in the coking region.

[0029] In this embodiment, by constructing a temperature matrix containing time and space dimensions on the heated surface and introducing a density reachability determination mechanism based on the temperature difference between measuring points, adaptive clustering identification of temperature anomalies is performed. This effectively suppresses the risk of misjudgment caused by local measurement noise and instantaneous temperature fluctuations, improving the spatial positioning accuracy and robustness of the actual coking high-temperature area. By combining the spatial range of the cluster with the centroid temperature for joint determination, the identification results not only reflect the intensity of the temperature anomaly but also reflect the scale characteristics of the anomaly area, enhancing the ability to distinguish coking areas with actual operational risks. By converting the flue gas flow field pressure signal from the time domain to the frequency domain and extracting the low-frequency dominant component, a stable pressure reference curve is constructed, effectively weakening the interference of combustion disturbance, turbulence pulsation, and measurement noise on the pressure signal. This makes the obtained pressure characteristics closer to the intrinsic variation law of the overall flow state of the boiler, improving the reliability of pressure information in characterizing flow field anomalies caused by coking.

[0030] In one alternative implementation, By combining pre-acquired measured pressure distribution data to determine the pressure deviation vector, and by combining the suspected coking area to obtain quantitative indicators of coking degree, including: Obtain the measured pressure distribution data at the time corresponding to the pressure reference curve, align the pressure reference curve and the measured pressure distribution data according to spatial coordinates, calculate the difference between the pressure value of the pressure reference curve at each spatial coordinate point and the pressure value of the measured pressure distribution data at the corresponding spatial coordinate point, and arrange them according to the spatial coordinate order to obtain the pressure deviation vector. Based on the spatial coordinates of the measured pressure distribution data, a spatial grid of the heated surface is constructed. The pressure deviation vector is mapped onto the spatial grid of the heated surface according to the spatial coordinates to form a pressure deviation distribution field. The spatial coordinate range of the suspected coking area is extracted. The pressure deviation values ​​of the grid nodes corresponding to the spatial coordinate range are extracted from the pressure deviation distribution field, and spatial gradient calculation is performed to obtain the pressure deviation gradient vector. Principal component analysis is performed on the pressure deviation gradient vector to obtain the dominant gradient direction and the variance contribution rate corresponding to each principal component. The angle between the dominant gradient direction and the flue gas flow direction is calculated to obtain the flow direction deviation. The variance contribution rate is used as a weight to perform a weighted summation of the pressure deviation values ​​of each grid node in the suspected coking area. The coking degree quantification index is obtained by combining the flow direction deviation and the spatial range of the suspected coking area.

[0031] After acquiring the pressure reference curve, the measured pressure distribution data at the corresponding time points are simultaneously collected. Taking the superheater area of ​​a 600 MW boiler as an example, an 8×6 pressure sensor grid with a total of 48 pressure measurement points is arranged. Each pressure measurement point records the real-time pressure value and its three-dimensional spatial coordinates. The pressure reference curve and the measured pressure distribution data are aligned according to spatial coordinates to ensure the spatiotemporal consistency of the data. A spatial interpolation algorithm is used during the alignment process to map data from different locations to a unified coordinate system.

[0032] The difference between the pressure value at each spatial coordinate point on the pressure reference curve and the pressure value at the corresponding spatial coordinate point in the measured pressure distribution data is calculated. For example, if the pressure reference curve value at a certain measuring point is 420 Pascals, and the corresponding measured pressure value is 385 Pascals, then the pressure deviation value at that point is 35 Pascals. The pressure deviation values ​​are arranged in order of spatial coordinates to form a pressure deviation vector. In the previous example, a pressure deviation vector containing 48 elements is obtained, with each element corresponding to the pressure deviation value of a pressure measuring point.

[0033] A spatial grid for the heated surface is constructed based on the spatial coordinates of measured pressure distribution data. The spatial grid is a three-dimensional structured grid with a size of 0.3 m × 0.3 m × 0.3 m, sufficiently detailed to represent the spatial variation characteristics of the pressure field. Pressure deviation vectors are mapped onto the heated surface spatial grid according to the spatial coordinates, forming a pressure deviation distribution field. The mapping process uses an inverse distance weighted interpolation method to ensure a smooth data transition. For each node in the grid, the corresponding pressure deviation value is obtained by a weighted average of the pressure deviation values ​​of surrounding measured points, with the weights inversely proportional to the distance.

[0034] The spatial coordinate range is extracted from the identified suspected coking areas. For example, the spatial coordinate range of a suspected coking area is 2.1 meters to 4.5 meters horizontally, 1.8 meters to 3.6 meters vertically, and 6.3 meters to 7.8 meters vertically. Within this spatial range, there are approximately 144 grid nodes. The pressure deviation values ​​of the grid nodes corresponding to this spatial coordinate range are extracted from the pressure deviation distribution field to form a local pressure deviation field.

[0035] Spatial gradient calculations are performed on the local pressure deviation field to obtain the pressure deviation gradient vector. The gradient calculation uses a central difference scheme. For internal nodes, the gradient components in each direction are calculated using the pressure deviation values ​​of the forward and backward nodes. For example, the gradient components of a certain node in the lateral, longitudinal, and height directions are 28 Pascals per meter, 15 Pascals per meter, and 42 Pascals per meter, respectively. For boundary nodes, a one-sided difference scheme is used for gradient calculation.

[0036] Principal component analysis (PCA) is performed on the pressure deviation gradient vector to determine the dominant gradient direction and the variance contribution rate of each principal component. During PCA, the gradient covariance matrix is ​​constructed, and then the eigenvalues ​​and eigenvectors are solved. The eigenvectors correspond to the principal component directions, and the magnitude of the eigenvalues ​​reflects the degree of data variation along that direction. In the previous example, three principal component directions were obtained: the first principal component direction is the vector (0.58, 0.31, 0.75), with a variance contribution rate of 68%; the second principal component direction is the vector (0.29, 0.91, -0.29), with a variance contribution rate of 24%; and the third principal component direction is the vector (0.76, -0.27, -0.59), with a variance contribution rate of 8%. The first principal component direction is the dominant gradient direction.

[0037] The angle between the dominant gradient direction and the boiler design flue gas flow direction is calculated to obtain the flow direction deviation. The design flue gas flow direction is usually obtained from the boiler design drawings; for example, the design flow direction for a certain superheater area is a vector (0.1, 0.2, 0.97). The angle between the two vectors is calculated using the vector dot product formula, yielding a flow direction deviation of 23.5 degrees. The larger the flow direction deviation, the more the local flue gas flow deviates from the design state, and the more significant the coking effect.

[0038] Using the variance contribution rate of each principal component as a weight, the pressure deviation values ​​of each grid node within the suspected coking region are weighted and summed. The weights reflect the degree of contribution of each direction to the variation in the pressure field. For example, if the pressure deviation value of a node is 40 Pascals, the weighted pressure deviation value after weighting by the variance contribution rates of the three principal components is 0.68×40 + 0.24×40 + 0.08×40 = 40 Pascals. A similar process is applied to all nodes within the region, and then the average is calculated, resulting in an average weighted pressure deviation value of 38.5 Pascals for the region.

[0039] A quantitative index for coking severity is constructed by combining the flow direction deviation, the spatial range of the suspected coking area, and the average weighted pressure deviation. The index is calculated as follows: average weighted pressure deviation multiplied by the ratio of the flow direction deviation to 90 degrees, then multiplied by the square root of the ratio of the suspected coking area volume to the reference volume. For example, if the flow direction deviation is 23.5 degrees, the suspected coking area volume is 3.78 cubic meters, the reference volume is set at 1 cubic meter, and the average weighted pressure deviation is 38.5 Pascals, then the quantitative index for coking severity is 38.5 × (1 + 23.5 / 90) × (3.78 / 1). 1 / 2 =62.4. This indicator comprehensively considers the degree of pressure anomaly, flow field distortion, and coking range, and can objectively reflect the severity of coking.

[0040] In this embodiment, by constructing a pressure deviation vector and mapping it to form a continuous pressure deviation distribution field, the pressure anomaly is elevated from a discrete measurement point to a spatially continuous expression. This effectively enhances the overall perception of spatial effects such as local blockage and flow channel contraction caused by coking, avoiding the problem that single-point pressure anomalies cannot reflect the true impact range of coking. Spatial gradient analysis of pressure deviation in suspected coking areas can reveal the directional characteristics of pressure changes, thereby reflecting the degree of disturbance of coking on flue gas flow path and pressure distribution structure. Principal component analysis is used to extract the dominant direction of the pressure deviation gradient, and the pressure deviation is weighted by the variance contribution rate. This makes the coking degree assessment more focused on the dominant change pattern that has a significant impact on the overall flow field structure, suppressing the interference of local random disturbances and measurement noise on the assessment results, and improving the stability and reliability of the coking quantification results. Combining the degree of deviation between the dominant gradient direction and the flue gas flow direction, the physical constraint of flow structure change is introduced, so that the coking degree index can directly reflect the degree of damage of coking to the flue gas flow organization, enhancing the correlation between the index and actual operational risks.

[0041] In one alternative implementation, Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. Heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the aforementioned coking degree quantification index. This includes: Historical boiler combustion conditions are obtained from the historical database, and the injection angle parameters and injection speed parameters of each burner are extracted from the historical boiler combustion conditions. Taking the spatial position of the burner as the starting point of the ray, the injection angle parameter determines the propagation direction of the ray, the injection speed parameter determines the propagation step size of the ray, and the ray is gradually tracked along the propagation direction and the intersection position of the ray with each spatial coordinate point of the heated surface is recorded. The cumulative number of times each spatial coordinate point is passed by the ray is counted and the heat load intensity value is determined. The heat load intensity values ​​of all spatial coordinate points are arranged according to the spatial coordinates to obtain the heat load intensity distribution. The spatial coordinate range of the suspected coking area is obtained. The heat load intensity value corresponding to the spatial coordinate range is extracted from the heat load intensity distribution and matched with the coking degree quantification index according to the spatial coordinates to obtain a paired dataset. The heat load intensity value in the paired dataset is divided into intervals to obtain the heat load intensity level. The coking degree quantification index in the paired dataset is divided into intervals to obtain the coking degree level. The heat load intensity level is used as the row index and the coking degree level is used as the column index. The coking risk assessment matrix is ​​constructed by traversing each spatial coordinate point in the paired dataset.

[0042] Historical boiler combustion data, spanning nearly 30 days, was obtained from the power plant's distributed control system historical database, with a sampling interval of one hour. Key parameters for each burner, including injection angle and injection velocity, were extracted from the historical boiler combustion data. Taking a 600 MW boiler as an example, this boiler is equipped with 16 burners, distributed in four burner layers around the furnace. Three parameters were recorded for each burner: horizontal injection angle, vertical injection angle, and injection velocity. For example, burner No. 1 had a horizontal injection angle of 15 degrees, a vertical injection angle of 5 degrees, and an injection velocity of 28 meters per second.

[0043] Fuel injection trajectory and heat load distribution are simulated using ray tracing. The spatial location of the burner is taken as the starting point of the ray. The propagation direction of the ray is determined by the injection angle parameter, and the propagation step size is determined by the injection velocity parameter. During ray tracing, the step size is set to 0.1 meters, and the ray advances step by step along the propagation direction. For example, if a burner is located at coordinates (2, 0, 10), with a horizontal injection angle of 20 degrees, a vertical injection angle of 10 degrees, and an injection velocity of 30 meters per second, then the starting point of the ray is (2, 0, 10), the propagation direction vector is (0.934, 0.342, 0.1736), and the corresponding single-step displacement is (0.0934, 0.0342, 0.01736) meters. At each step forward, the current ray position is recorded, and it is determined whether this position intersects with the boiler heating surface.

[0044] During ray tracing, the boiler heating surface is discretized into a spatial grid with a grid size of 0.2 m × 0.2 m × 0.2 m. For each ray, all grid cells it passes through are recorded. When a ray intersects a grid cell on the heating surface, the count of that grid cell is incremented. For example, a grid cell initially has a count of 0, and after passing through 10 rays, the count increases to 10. After all ray tracing is completed, the cumulative number of times each grid cell is traversed by rays is counted, and the heat load intensity value is determined based on the cumulative number of traversals. The heat load intensity value is calculated by considering ray density and fuel combustion characteristics, and is obtained by multiplying the cumulative number of traversals by the unit fuel calorific value and dividing by the grid volume, with the unit being kilowatts per cubic meter.

[0045] The heat load intensity values ​​of all spatial coordinate points are arranged according to their spatial coordinates to form a three-dimensional heat load intensity distribution matrix. The dimensions of this matrix are consistent with the spatial grid of the boiler's heating surface, with each element corresponding to the heat load intensity value of a grid cell. In the actual case, the boiler's heating surface spatial grid is divided into a 100×80×120 three-dimensional grid, totaling 960,000 grid cells. Each cell stores a heat load intensity value, typically ranging from 0 to 500 kW per cubic meter. For example, the grid cell at coordinates (50, 40, 60) has a heat load intensity value of 320 kW per cubic meter.

[0046] Obtain the spatial coordinate range of the suspected coking area. For example, the spatial range of a suspected coking area is horizontal index 45 to 55, vertical index 35 to 45, and height index 55 to 65, containing a total of 11×11×11=1331 grid cells. Extract the heat load intensity values ​​corresponding to this spatial coordinate range from the heat load intensity distribution matrix to form a local heat load intensity distribution submatrix.

[0047] The local heat load intensity distribution and coking degree quantification index were matched according to spatial coordinates to obtain a paired dataset. Each record in the paired dataset contains three key pieces of information: spatial coordinates, heat load intensity value, and coking degree quantification index. For example, the grid cell at coordinates (48, 38, 58) has a heat load intensity value of 285 kW per cubic meter, and the corresponding coking degree quantification index is 52.3. Similar matching processing was performed on all 1331 grid cells in the entire suspected coking area to form a paired dataset containing 1331 records.

[0048] The heat load intensity values ​​in the paired dataset were divided into intervals to obtain heat load intensity levels. The interval division employed an equal-frequency binning method, classifying the heat load intensity values ​​into five levels: extremely low (0-100 kW / m³), low (100-200 kW / m³), medium (200-300 kW / m³), high (300-400 kW / m³), and extremely high (400-500 kW / m³). Similarly, the coking severity quantification index was divided into intervals to obtain coking severity levels. The coking severity quantification index was classified into four levels: slight (0-30), moderate (30-60), severe (60-90), and extremely severe (above 90).

[0049] A coking risk assessment matrix was constructed using heat load intensity level as the row index and coking severity level as the column index. The matrix has a dimension of 5×4 and a total of 20 cells. All records in the paired dataset were traversed, and each record was assigned to its corresponding cell based on its heat load intensity level and coking severity level, with the number of records counted. For example, there were 328 grid cells with a heat load intensity of "medium" and a coking severity of "moderate," accounting for 24.6% of the total sample. The resulting coking risk assessment matrix is ​​as follows: 180 points (13.5%) for "extremely low" heat load and "slight" coking; 65 points (4.9%) for "extremely low" heat load and "moderate" coking; 42 points (3.2%) for "low" heat load and "severe" coking; and 95 points (7.1%) for "high" heat load and "extremely severe" coking.

[0050] In this embodiment, a ray propagation model is constructed based on the burner injection angle and injection velocity to statistically analyze the heat load intensity at various spatial locations of the heated surface, forming a heat load intensity distribution directly related to the combustion structure. This allows the coking risk assessment to reflect the long-term impact of the combustion organization on the heat input of the heated surface, avoiding spatial distortion caused by relying solely on experience or uniform assumptions about heat load. This improves the spatial resolution and physical rationality of the heat load characterization. By matching the heat load intensity and coking degree quantification indicators in the suspected coking area one-to-one at the spatial coordinate level and constructing a coking risk assessment matrix at different levels, the coupled expression of heat load driving factors and coking result characteristics is realized, which is beneficial for distinguishing different risk types and improving the precision of risk judgment.

[0051] In one alternative implementation, The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed using pre-acquired flue gas flow paths. Based on the directed weighted graph, the coking propagation coefficient is calculated, and coking acceleration nodes are determined using Markov chains. Based on the coking acceleration nodes, a sootblower activation sequence is determined, and a time interval allocation scheme is generated using the coking propagation coefficient, including: The elements in the coking risk assessment matrix are sorted in descending order according to their numerical values ​​to obtain an element sequence. The dependencies between different matrix elements are determined based on the heat load intensity level and coking degree level corresponding to each element, and the element sequence is topologically sorted to obtain a priority sequence. Project the spatial coordinate points corresponding to the elements in the priority sequence along the flue gas flow path to obtain the projection position. Use the spatial coordinate points as graph nodes, determine the adjacent relationship between the graph nodes according to the projection position as the edge connection relationship, and construct a directed weighted graph using the elements of each spatial coordinate point in the coking risk assessment matrix as the node weight value. The outgoing edge propagation intensity is obtained by traversing the directed weighted graph and calculating the ratio of the weight value corresponding to the current graph node to the weight value corresponding to the downstream node. The outgoing edge propagation intensity of each graph node is summed to obtain the coking propagation coefficient. A Markov chain is constructed with the graph nodes as states and the coking propagation coefficient as the state transition probability, and a steady-state probability distribution is obtained by iterative calculation. Based on the steady-state probability distribution, coking acceleration nodes are determined. The spatial coordinates of each coking acceleration node are extracted and the corresponding sootblower number is queried. The sootblower start sequence is determined according to the order of the coking acceleration nodes on the flue gas flow path and the sootblower number. The time interval is determined by combining the coking propagation coefficient between adjacent coking acceleration nodes and a time interval allocation scheme is generated.

[0052] A thorough analysis was conducted on the obtained coking risk assessment matrix, a 5×4 two-dimensional array where rows represent heat load intensity levels and columns represent coking severity levels. The elements were sorted in descending order of their numerical values ​​to obtain an element sequence. Taking a 600 MW boiler as an example, the first five elements after sorting are: "Extremely High Heat Load" - "Extremely Severe Coking", value 325; "High Heat Load" - "Severe Coking", value 287; "Extremely High Heat Load" - "Severe Coking", value 246; "High Heat Load" - "Extremely Severe Coking", value 231; "Medium Heat Load" - "Extremely Severe Coking", value 204. These elements represent the operating condition combinations with the highest coking risk and require priority handling.

[0053] Based on the corresponding heat load intensity level and coking degree level of each element, the dependencies between different matrix elements are determined. The determination of these dependencies follows these rules: when the heat load intensity level is the same, elements with a higher degree of coking depend on elements with a lower degree of coking; when the coking degree level is the same, elements with a higher heat load intensity depend on elements with a lower heat load intensity. For example, the element with "extremely high" heat load and "extremely severe" coking depends on the element with "extremely high" heat load and "severe" coking, which in turn depends on the element with "extremely high" heat load and "moderate" coking. This dependency indicates that coking typically develops gradually from a mild state to a severe state, and that heat load promotes coking development.

[0054] Based on the established dependencies, the element sequence is topologically sorted to obtain a priority sequence. Topological sorting ensures that dependent nodes are processed before dependent nodes. The sorting algorithm uses depth-first search, starting from the leaf nodes of the dependency graph and progressively building the sorted sequence upwards. For example, the first few elements of a priority sequence might be: "Low" heat load - "Slight" coking, "Medium" heat load - "Slight" coking, "Low" heat load - "Moderate" coking, "Medium" heat load - "Moderate" coking, and "High" heat load - "Slight" coking.

[0055] The spatial coordinates corresponding to the elements in the priority sequence are projected along the flue gas flow path to obtain the projected position. The flue gas flow path is obtained through the boiler's computational fluid dynamics model, which considers factors such as boiler geometry, combustion conditions, and heating surface arrangement to calculate the three-dimensional flow field distribution inside the boiler. During the projection process, the spatial coordinates are projected downstream along the flue gas flow direction at that point by a certain distance, for example, 2 meters, to obtain the projected position. For a certain element in the priority sequence, its corresponding spatial coordinates are (42, 35, 68), the flue gas flow direction at that point is (0.2, 0.1, 0.97), and the projected position is (42.4, 35.2, 69.94).

[0056] Using spatial coordinate points as graph nodes, the adjacency relationships between nodes are determined based on their projected positions, forming edge connections. If the projected position of one node is sufficiently close to the spatial position of another node (e.g., less than 1 meter), a directed edge is established between these two nodes, pointing from the upstream node to the downstream node. The element values ​​of each spatial coordinate point in the coking risk assessment matrix are used as node weight values ​​to construct a directed weighted graph. For example, the node with spatial coordinates (42, 35, 68) has a weight value of 245, and a directed edge is established between this node and the downstream node (43, 35, 70), whose node weight value is 187.

[0057] Traverse the directed weighted graph and calculate the ratio of the weight value of the current node to the weight value of the downstream node to obtain the outgoing edge propagation strength. For example, if the weight value of node (42, 35, 68) is 245, and the weight value of the corresponding downstream node (43, 35, 70) is 187, then the propagation strength of this outgoing edge is 245 divided by 187, approximately 1.31. If a node has multiple outgoing edges, calculate the propagation strength of each outgoing edge separately. Sum the outgoing edge propagation strengths of each graph node to obtain the coking propagation coefficient. For example, if node (42, 35, 68) has 3 outgoing edges with propagation strengths of 1.31, 1.22, and 1.17 respectively, then the coking propagation coefficient of this node is 3.7. The coking propagation coefficient reflects the ability of the coking state at this point to propagate downstream; the larger the coefficient, the more significant the impact of the coking state at this point on the downstream.

[0058] Construct a Markov chain with graph nodes as states and cohesion propagation coefficients as state transition probabilities. The state transition probability matrix is ​​constructed as follows: for each edge in the directed graph, exemplified by a path from node A to node B, the probability of transitioning from state A to state B is set to the outgoing edge propagation strength of node A divided by the cohesion propagation coefficient of node A. For example, the probability of transitioning from state (42, 35, 68) to state (43, 35, 70) is 1.31 divided by 3.7, approximately 0.35. Iterate the Markov chain until a steady-state probability distribution is reached. The specific iteration process uses a power-law iteration method, with an initial uniformly distributed state vector, and the number of iterations is no less than 100 or until the difference between two adjacent iterations is less than a preset threshold of 0.001.

[0059] Based on the steady-state probability distribution, coking acceleration nodes were identified. Coking acceleration nodes are those whose steady-state probability values ​​exceed a preset threshold, such as 0.02. These nodes are key points in the coking propagation process and have a significant impact on the formation and development of coking. For a suspected coking area in a boiler, 15 coking acceleration nodes were identified. The coordinates of the top three nodes in terms of steady-state probability were (42, 35, 68), (48, 37, 72), and (44, 34, 75), with steady-state probability values ​​of 0.058, 0.047, and 0.043, respectively.

[0060] Extract the spatial coordinates of each coking acceleration node and query the corresponding sootblower number. The spatial arrangement information of the sootblowers in the boiler is stored in the equipment database. A spatial location matching algorithm is used to determine the nearest sootblower for each coking acceleration node. The matching algorithm calculates the spatial distance between each coking acceleration node and each sootblower, and selects the sootblower with the smallest distance and less than a preset threshold, such as 3 meters, as the processing device for that node. For example, the nearest sootblower number for coking acceleration node (42, 35, 68) is number 7, the nearest sootblower number for node (48, 37, 72) is number 12, and the nearest sootblower number for node (44, 34, 75) is number 9.

[0061] The sootblower start-up sequence is determined based on the order of the coking acceleration nodes along the flue gas flow path and their sootblower numbers. The principle for determining the start-up sequence is to proceed step-by-step from upstream to downstream; that is, the upstream coking acceleration nodes are processed first, followed by the downstream nodes. The node processing order is obtained by sorting the coking acceleration nodes according to their height coordinates (representing the main direction of flue gas flow). For example, the sootblower start-up sequence corresponding to the first 5 sorted coking acceleration nodes is 7, 5, 12, 9, and 14.

[0062] By combining the coking propagation coefficient between adjacent coking acceleration nodes, the time interval for starting the sootblowers is determined, generating a time interval allocation scheme. The calculation of the time interval considers two factors: the coking propagation coefficient and the spatial distance between nodes. A larger coking propagation coefficient indicates a more rapid spread of coking effects, requiring a shorter time interval; a larger spatial distance between nodes requires a longer time interval to ensure that upstream sootblowing is fully utilized before downstream sootblowing. For adjacent sootblowers No. 7 and No. 5, since the coking propagation coefficient between the corresponding nodes is 3.2 and the spatial distance is 4.5 meters, the calculated time interval is 15 minutes; for sootblowers No. 5 and No. 12, the time interval is 20 minutes; for sootblowers No. 12 and No. 9, the time interval is 12 minutes; and for sootblowers No. 9 and No. 14, the time interval is 18 minutes.

[0063] In this embodiment, topological sorting is achieved by ranking the elements of the coking risk assessment matrix and introducing dependency constraints. This allows high-risk areas with more significant downstream impacts to receive higher priority in decision-making, avoiding the problem of unreasonable soot blowing order caused by ignoring spatial correlation and sequential impact relationships based solely on risk magnitude. This improves the global rationality of soot blowing resource scheduling. By mapping coking risk to a directed weighted graph structure along the flue gas flow path and characterizing the downstream diffusion of coking impact with coking propagation intensity, coking assessment is upgraded from a static spatial distribution to a dynamic structural expression that can describe the propagation trend. This enhances the ability to characterize the coking development direction and diffusion potential. Based on the steady-state probability identification of Markov chain, coking acceleration nodes are identified, making the determination of key coking propagation nodes globally convergent and stable, reducing the impact of instantaneous disturbances or local anomalies on the key node identification results.

[0064] Figure 2 This is a flowchart illustrating the coking risk assessment and sootblower scheduling optimization process of the sootblowing control optimization method for preventing coking in thermal power unit boilers, as described in an embodiment of the present invention.

[0065] In one alternative implementation, Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the sootblower start-up sequence, calculate the duration of operation, including: The steam pipeline network monitoring system acquires steam pipeline network pressure fluctuation data within a preset time period and performs time-domain decomposition to obtain a pressure reference component and a pressure fluctuation component. Spectral analysis is performed on the pressure fluctuation component to obtain the dominant frequency component and its corresponding amplitude. Based on the amplitude corresponding to the dominant frequency component, the pressure fluctuation intensity is calculated using an energy accumulation method. The steady-state available steam flow rate is calculated based on the pressure reference component. The flow loss corresponding to the safety margin is calculated based on the pressure fluctuation intensity to obtain the fluctuation margin. Based on the steady-state available steam flow rate and the fluctuation margin, the upper limit of the available steam flow rate is obtained using a margin deduction method. Extract the number of each sootblower in the sootblower start-up sequence and query the corresponding rated steam consumption and rated operating time. Obtain the coking propagation coefficient of the coking acceleration node corresponding to each sootblower and perform normalization processing to obtain the coking propagation weight corresponding to each sootblower. Based on the rated steam consumption and the coking propagation weight, obtain the weighted steam consumption by weighted summation. Determine the sufficiency of the available steam flow rate by comparing the upper limit of available steam flow rate with the weighted steam consumption to obtain the determination result. Based on the determination result, combine the rated operating time and the coking propagation weight to solve for the operating duration of each sootblower.

[0066] Steam network pressure fluctuation data for a preset time period (approximately 4 hours) was acquired from the power plant's steam network monitoring system, with a sampling frequency of 10 Hz. Taking a 600 MW unit as an example, 144,000 steam network pressure data points were collected, with pressure values ​​ranging from 3.8 to 4.2 MPa. The acquired pressure fluctuation data underwent time-domain decomposition using the empirical mode decomposition method, decomposing the original pressure signal into multiple intrinsic mode functions (IMFs) and a residual term. The residual term serves as the pressure reference component, representing the average pressure level of the steam network; the superposition of the IMFs serves as the pressure fluctuation component, representing the dynamic characteristics of pressure changes. For example, the decomposed pressure reference component was 4.05 MPa, and the peak value range of the pressure fluctuation component was ±0.15 MPa.

[0067] Spectral analysis was performed on the pressure fluctuation components using a Fast Fourier Transform (FFT) algorithm to convert the time-domain signal into frequency-domain features. The frequency resolution was set to 0.01 Hz, and the highest analysis frequency was 5 Hz. The dominant frequency components and their corresponding amplitudes were identified by amplitude sorting. In this implementation, the identified dominant frequency components were 0.15 Hz, 0.32 Hz, 0.48 Hz, and 0.64 Hz, with corresponding amplitudes of 0.08 MPa, 0.06 MPa, 0.04 MPa, and 0.03 MPa, respectively. The dominant frequency components reflect the main periodic characteristics of the steam pipeline network pressure fluctuations and are related to system operating conditions, equipment operating characteristics, and external disturbances.

[0068] The pressure fluctuation intensity is calculated based on the amplitude corresponding to the dominant frequency component using an energy accumulation method. The normalized pressure fluctuation intensity index is obtained by dividing the sum of the squares of the amplitudes of each dominant frequency component by the total spectral energy. During the calculation, frequency weighting is applied, with higher frequency fluctuations having a greater weight than lower frequency fluctuations. In this embodiment, the calculated pressure fluctuation intensity is 0.32, indicating that the pressure fluctuation energy accounts for 32% of the total energy.

[0069] The steady-state available steam flow rate is calculated based on the pressure reference component. The relationship between pressure and flow rate considers factors such as pressure, temperature, and pipeline characteristics. For a pressure reference component of 4.05 MPa, the corresponding steady-state available steam flow rate is 28.4 tons per hour. The flow loss corresponding to the safety margin is calculated based on the pressure fluctuation intensity, yielding the fluctuation margin. The fluctuation margin calculation considers the pressure fluctuation intensity, system safety factor, and flow response characteristics. When the pressure fluctuation intensity is 0.32, the corresponding flow loss is 5.7 tons per hour. Based on the steady-state available steam flow rate and the fluctuation margin, the upper limit of the available steam flow rate is obtained through the margin deduction method. Subtracting the fluctuation margin from the steady-state available steam flow rate yields the maximum available steam flow rate under the condition of ensuring stable system operation. In this example, the upper limit of the available steam flow rate is 22.7 tons per hour.

[0070] Extract the numbers of each sootblower in the aforementioned sootblower start-up sequence. In this embodiment, this includes sootblowers No. 7, No. 5, No. 12, No. 9, and No. 14. Query the rated steam consumption and rated operating time for each sootblower. For example, sootblower No. 7 has a rated steam consumption of 4.2 tons per hour and a rated operating time of 40 minutes; sootblower No. 5 has a rated steam consumption of 3.8 tons per hour and a rated operating time of 35 minutes; sootblower No. 12 has a rated steam consumption of 4.5 tons per hour and a rated operating time of 42 minutes; sootblower No. 9 has a rated steam consumption of 4.0 tons per hour and a rated operating time of 38 minutes; and sootblower No. 14 has a rated steam consumption of 3.9 tons per hour and a rated operating time of 36 minutes.

[0071] Obtain the coking propagation coefficients for the coking acceleration nodes corresponding to each sootblower. For example, the coking propagation coefficient for sootblower No. 7 is 3.7, for sootblower No. 5 it is 3.2, for sootblower No. 12 it is 2.9, for sootblower No. 9 it is 3.4, and for sootblower No. 14 it is 3.0. Normalize these coking propagation coefficients to obtain the coking propagation weights for each sootblower. The normalization process divides each coking propagation coefficient by its sum to ensure that the sum of the weights is 1. After normalization, the coking propagation weight for sootblower No. 7 is 0.23, for sootblower No. 5 it is 0.20, for sootblower No. 12 it is 0.18, for sootblower No. 9 it is 0.21, and for sootblower No. 14 it is 0.18.

[0072] Based on the rated steam consumption and coking propagation weight, the weighted steam consumption is obtained through weighted summation. The calculation method involves multiplying the rated steam consumption of each sootblower by its coking propagation weight and then summing the results. In this embodiment, the weighted steam consumption is 4.09 tons per hour. The available steam flow rate is then compared with the weighted steam consumption to determine sufficiency. The determination method involves comparing the magnitude of the available steam flow rate and the weighted steam consumption. In this embodiment, the available steam flow rate is 22.7 tons per hour, and the weighted steam consumption is 4.09 tons per hour; the determination result is that the steam flow rate is sufficient.

[0073] Based on the judgment results, and combined with the rated operating time and coking propagation weight, the operating duration of each sootblower is calculated. When the steam flow is sufficient, the operating duration of each sootblower can be optimally allocated based on the corresponding rated operating time and coking propagation weight. The calculation method is to multiply the rated operating time by an adjustment coefficient, which is proportional to the coking propagation weight. In specific implementation, the adjustment coefficient is set as the coking propagation weight divided by the average weight. In this embodiment, the average weight is 0.2, so the adjustment coefficient for sootblower No. 7 is 1.15, corresponding to an operating duration of 46 minutes; the adjustment coefficient for sootblower No. 5 is 1.0, corresponding to an operating duration of 35 minutes; the adjustment coefficient for sootblower No. 12 is 0.9, corresponding to an operating duration of 38 minutes; the adjustment coefficient for sootblower No. 9 is 1.05, corresponding to an operating duration of 40 minutes; and the adjustment coefficient for sootblower No. 14 is 0.9, corresponding to an operating duration of 32 minutes.

[0074] If the steam flow is insufficient, a constrained optimization method is needed to determine the duration of operation. The optimization objective is to maximize the coking removal effect, with the constraint that the total steam consumption does not exceed the upper limit of the available steam flow. The optimization problem is solved using the Lagrange multiplier method to obtain the duration of operation for each sootblower. For example, under insufficient steam flow conditions, the duration of operation for sootblower #7 is adjusted to 36 minutes, for sootblower #5 to 30 minutes, for sootblower #12 to 32 minutes, for sootblower #9 to 34 minutes, and for sootblower #14 to 28 minutes.

[0075] For extreme situations involving severe steam flow shortages, an emergency response strategy is implemented. Sootblowers are prioritized based on their coking propagation weight, ensuring the operational duration of high-weight sootblowers is guaranteed, while low-weight sootblowers can be temporarily shut down or have their operational duration reduced. For example, when the maximum available steam flow is below 15 tons per hour, sootblower number 14 can be temporarily shut down, allocating the limited steam resources to the other four sootblowers.

[0076] Based on the aforementioned calculations, a detailed sequence of sootblowing control instructions is generated. Each instruction includes the sootblower number, start time, duration of operation, and steam flow control parameters. For example, sootblower number 7 starts at time t, operates for 46 minutes, and has a steam flow rate of 4.2 tons per hour; sootblower number 5 starts at t+15 minutes, operates for 35 minutes, and has a steam flow rate of 3.8 tons per hour; and so on, completing the entire sootblowing operation sequence.

[0077] In this embodiment, by performing time-domain decomposition and spectral analysis on the pressure fluctuations of the steam pipeline network, the steady-state pressure level and the energy intensity of the fluctuations are distinguished. Based on the pressure fluctuation intensity, a safety margin is reserved to determine the upper limit of the available steam flow rate. This effectively avoids the problem of drastic pressure fluctuations in the pipeline network caused by relying solely on rated pressure or instantaneous flow rate for soot blowing decisions in traditional methods. This improves the overall stability and anti-disturbance capability of the steam pipeline network operation. By introducing a coking propagation weight to weighted evaluate the steam consumption of each soot blower, the allocation of steam resources prioritizes soot blowers that have a more significant effect on inhibiting coking diffusion. This avoids the ineffective consumption of steam resources in low-yield areas and improves the overall effectiveness of soot blowing under limited steam supply conditions. Combined with the flow sufficiency judgment results, the duration of each soot blower's action is adaptively adjusted so that the soot blowing intensity can achieve a dynamic balance between coking risk control and steam supply capacity.

[0078] In one alternative implementation, Based on the duration of the action and the time interval allocation scheme, the total execution time is determined and it is judged whether it exceeds the time window constraint. If it does, the duration is compressed proportionally to obtain a corrected duration, and combined with the soot blower start sequence, a soot blowing control command sequence is generated and issued for execution, including: The time interval between adjacent sootblowers in the sootblower start-up sequence is extracted from the time interval allocation scheme and accumulated with the duration of action to obtain the total execution time. The current boiler operating condition data is obtained and the current load status and flue gas temperature are extracted. The corresponding time window constraint upper limit is queried according to the current load status and the flue gas temperature. It is determined whether the total execution time is greater than the time window constraint upper limit. If it is greater, the ratio coefficient between the total execution time and the time window constraint upper limit is calculated. The compression operation is performed based on the duration of action and the ratio coefficient to obtain the corrected duration. The steady-state probability value of the coking acceleration node corresponding to each sootblower is extracted and the priority weight is calculated. Based on the priority weight, the correction duration is redistributed and adjusted to obtain the optimal duration. Based on the optimal duration and the sootblower number in the sootblower start sequence, the start time and stop time of each sootblower are generated sequentially and encapsulated to obtain sootblowing control instructions. The sootblowing control instructions are arranged in the order of start time to generate a sootblowing control instruction sequence and sent to the sootblower control unit for execution through the industrial control network.

[0079] The time intervals between adjacent sootblowers in the aforementioned time interval allocation scheme are extracted. Taking a 600 MW unit as an example, the sootblower start-up sequence includes five sootblowers, numbered 7, 5, 12, 9, and 14, with time intervals of 15 minutes, 20 minutes, 12 minutes, and 18 minutes between adjacent sootblowers, respectively. Combining this with the previously determined duration of each sootblower's operation, sootblower 7 operates for 46 minutes, sootblower 5 for 35 minutes, sootblower 12 for 38 minutes, sootblower 9 for 40 minutes, and sootblower 14 for 32 minutes. The total execution time is calculated by adding the time intervals to the duration of operation. The calculation method starts from the start time of the first sootblower and ends at the finish time of the last sootblower, including the duration of operation of all sootblowers and the time intervals between adjacent sootblowers. In this embodiment, the total execution time is 46 minutes + 15 minutes + 35 minutes + 20 minutes + 38 minutes + 12 minutes + 40 minutes + 18 minutes + 32 minutes, which is calculated to be 236 minutes, or 3 hours and 56 minutes.

[0080] The system acquires current boiler operating condition data, including boiler load, main steam parameters, combustion conditions, and flue gas characteristics. The current load state and flue gas temperature are extracted as key parameters for assessing boiler coking status and determining the soot blowing time window. In this embodiment, the acquired current load state is 85% of rated load, and the flue gas temperature is 138 degrees Celsius. Based on the current load state and flue gas temperature, the corresponding upper limit of the time window constraint is queried. The upper limit of the time window constraint is a parameter preset based on boiler operating experience and coking dynamic characteristics, stored in the operating condition constraint database. The query method uses a two-dimensional interpolation method, using the load state and flue gas temperature as indexes to find the best matching time window constraint value. For the combination of 85% load state and 138 degrees Celsius flue gas temperature, the queried upper limit of the time window constraint is 180 minutes, or 3 hours.

[0081] The system determines whether the total execution time exceeds the upper limit of the time window constraint. In this embodiment, the total execution time is 236 minutes, and the upper limit of the time window constraint is 180 minutes. The result is that the total execution time exceeds the upper limit of the time window constraint. Since the total execution time exceeds the time window constraint, the execution plan needs to be adjusted. The ratio coefficient between the total execution time and the upper limit of the time window constraint is calculated, i.e., 236 minutes divided by 180 minutes, resulting in a ratio coefficient of 1.31. Based on the duration of each sootblower's action and the calculated ratio coefficient, a compression operation is performed to obtain the corrected duration. The compression operation uses a proportional reduction method, dividing the duration of each sootblower's action by the ratio coefficient. The calculated corrected durations are: sootblower #7: 35 minutes; sootblower #5: 27 minutes; sootblower #12: 29 minutes; sootblower #9: 31 minutes; and sootblower #14: 24 minutes.

[0082] The steady-state probability values ​​of the coking acceleration nodes corresponding to each sootblower were extracted, and priority weights were calculated. The steady-state probability values ​​were derived from the aforementioned Markov chain analysis results: the steady-state probability value for the coking acceleration node corresponding to sootblower #7 was 0.058, for sootblower #5 it was 0.042, for sootblower #12 it was 0.047, for sootblower #9 it was 0.043, and for sootblower #14 it was 0.039. The priority weights were calculated by dividing the steady-state probability value of each node by the sum of the steady-state probability values ​​of all nodes, ensuring that the sum of the weights was 1. The calculated priority weights were 0.25 for sootblower #7, 0.18 for sootblower #5, 0.21 for sootblower #12, 0.19 for sootblower #9, and 0.17 for sootblower #14.

[0083] The optimal duration of the correction is obtained by redistributing the time based on priority weights. The redistribution method allocates the total available time according to priority weights, while also considering the minimum effective operating time requirement of each sootblower. The total available time (excluding the time interval between adjacent sootblowers) is multiplied by the priority weight of each sootblower to obtain the allocated time. In this embodiment, the total available time is 180 minutes minus all time intervals (65 minutes in total), which is 115 minutes. Based on the priority weights, sootblower #7 is allocated 28.75 minutes, sootblower #5 20.7 minutes, sootblower #12 24.15 minutes, sootblower #9 21.85 minutes, and sootblower #14 19.55 minutes. Considering the integer time requirements of actual operation, the allocation results are rounded to obtain the optimal duration: 29 minutes for soot blower No. 7, 21 minutes for soot blower No. 5, 24 minutes for soot blower No. 12, 22 minutes for soot blower No. 9, and 20 minutes for soot blower No. 14.

[0084] Based on the optimal duration and the sootblower numbers in the sootblower start-up sequence, the start and stop times of each sootblower are generated sequentially and encapsulated to obtain sootblowing control instructions. Assuming the initial execution time is t, the start time of sootblower #7 is t, and the stop time is t+29 minutes; the start time of sootblower #5 is t+15 minutes, and the stop time is t+36 minutes; the start time of sootblower #12 is t+35 minutes, and the stop time is t+59 minutes; the start time of sootblower #9 is t+47 minutes, and the stop time is t+69 minutes; and the start time of sootblower #14 is t+65 minutes, and the stop time is t+85 minutes. The encapsulation of the sootblowing control instructions adopts the standard format of the industrial control protocol, including information such as device identification, operation type, execution time, duration, and control parameters. For example, the control instruction for sootblower #7 includes the device identification "Sootblower #7", operation type "Start", execution time "t", duration "29 minutes", and control parameter "Steam pressure 4.2 MPa".

[0085] The sootblowing control commands are arranged in chronological order of their start times to generate a sootblowing control command sequence. The sequence is ordered from earliest to latest according to the start time of the sootblowers to ensure that the control system can execute each operation in the correct timing. The ordered sootblowing control command sequence is as follows: Sootblower No. 7 start command (t), Sootblower No. 7 stop command (t+29 minutes), Sootblower No. 5 start command (t+15 minutes), Sootblower No. 5 stop command (t+36 minutes), Sootblower No. 12 start command (t+35 minutes), Sootblower No. 12 stop command (t+59 minutes), Sootblower No. 9 start command (t+47 minutes), Sootblower No. 9 stop command (t+69 minutes), Sootblower No. 14 start command (t+65 minutes), Sootblower No. 14 stop command (t+85 minutes).

[0086] The sootblowing control command sequence is sent to the sootblower control unit via an industrial control network. The industrial control network employs redundant fieldbus technology to ensure reliable and real-time communication. Data transmission is encrypted to prevent command tampering. After the control command is sent, the sootblower control unit automatically executes the sootblowing operation according to the timing requirements of the command, without manual intervention. Simultaneously, the system monitors various parameters during the sootblowing process in real time, including steam pressure, steam flow rate, and sootblower movement status, ensuring the sootblowing operation proceeds as expected. If an abnormality is detected, such as a sudden drop in steam pressure or sootblower jamming, the control system automatically interrupts the current operation and issues an alarm to prevent equipment damage and safety accidents.

[0087] In this embodiment, by comparing the overall execution time of the sootblower with the time window constraints determined based on the current load state and flue gas temperature, and uniformly compressing the duration of each sootblower's action when the limits are exceeded, the problem of the sootblowing process crossing unfavorable operating ranges or conflicting with key operating condition adjustment stages is effectively avoided. This improves the coordination between sootblowing operations and boiler operating rhythm, reduces the adverse impact on the stable operation of the unit, and by introducing the steady-state probability of the coking acceleration node as a priority weight, the compressed duration is redistributed so that the limited execution time resources are preferentially allocated to sootblowers that have a more significant effect on inhibiting coking propagation. This avoids the problem of insufficient sootblowing in key areas due to simple proportional compression, and maintains the overall optimality of sootblowing effect within the limited time window. By generating a sootblowing control command sequence containing precise start and stop times and issuing and executing it in an orderly manner, deterministic scheduling and traceable control of sootblowing actions are realized, reducing the need for manual intervention and temporary adjustments, and improving the consistency and reliability of control execution.

[0088] A second aspect of the present invention provides a soot blowing control optimization system for preventing coking in thermal power unit boilers, comprising: The coking sensing module is used to collect boiler heating surface temperature data and flue gas flow field pressure distribution data. It performs spatiotemporal clustering analysis on the heating surface temperature data to identify suspected coking areas, performs Fourier transform on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve, determines the pressure deviation vector by combining the pre-acquired measured pressure distribution data, and obtains the coking degree quantification index by combining the suspected coking areas. The risk assessment module is used to obtain historical boiler combustion conditions and extract burner injection angle parameters, calculate heat load intensity distribution through ray tracing algorithm, and construct a coking risk assessment matrix by combining the coking degree quantification index. The sequence generation module is used to perform topological sorting on the coking risk assessment matrix and construct a directed weighted graph in combination with the pre-acquired flue gas flow path. Based on the directed weighted graph, the coking propagation coefficient is calculated and the coking acceleration node is determined through a Markov chain. Based on the coking acceleration node, the soot blower start sequence is determined and a time interval allocation scheme is generated in combination with the coking propagation coefficient. The soot blowing control module is used to acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the soot blower start sequence, it calculates the duration of action. Based on the duration of action and the time interval allocation scheme, it determines the total execution time and determines whether it exceeds the time window constraint. If it exceeds the constraint, it compresses the duration proportionally to obtain a corrected duration and combines it with the soot blower start sequence to generate a soot blowing control command sequence for execution.

[0089] A third aspect of the present invention provides an electronic device, comprising: A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0090] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0091] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimized method for soot blowing control to prevent coking in thermal power unit boilers, characterized in that, include: The boiler's heating surface temperature data and flue gas flow field pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve. The pressure deviation vector is determined by combining the pre-acquired measured pressure distribution data. The coking degree quantification index is obtained by combining the suspected coking areas. Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. The heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the coking degree quantification index. The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed by combining it with the pre-acquired flue gas flow path. The coking propagation coefficient is calculated based on the directed weighted graph and the coking acceleration node is determined by the Markov chain. The soot blower start sequence is determined based on the coking acceleration node and a time interval allocation scheme is generated by combining it with the coking propagation coefficient. Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Calculate the duration of action based on the upper limit of available steam flow and the sootblower start-up sequence. Determine the total execution time based on the duration of action and the time interval allocation scheme, and determine whether it exceeds the time window constraint. If it does, compress the duration proportionally to obtain a corrected duration and combine it with the sootblower start-up sequence to generate a sootblowing control command sequence for execution.

2. The method according to claim 1, characterized in that, The boiler's heating surface temperature data and flue gas flow pressure distribution data are collected. Spatiotemporal clustering analysis is performed on the heating surface temperature data to identify suspected coking areas. Fourier transform is performed on the flue gas flow pressure distribution data to reconstruct the pressure baseline curve, including: The distributed temperature sensor array collects multi-point temperature data of the boiler heating surface at a preset sampling frequency, and the pressure sensor array collects multi-point pressure data of the flue gas flow field at a preset sampling frequency. Based on the timestamps and spatial coordinates corresponding to the multi-point temperature data, a spatiotemporal temperature matrix is ​​constructed. The temperature difference between each measuring point in the spatiotemporal temperature matrix and its adjacent measuring points is calculated. The temperature difference is used as a distance metric to determine the density reachability of the measuring points. The density-reachable measuring points are merged into multiple temperature clusters. The centroid temperature and spatial range of each temperature cluster are calculated. It is determined whether the centroid temperature is higher than a preset temperature threshold and whether the spatial range is greater than a preset range threshold. If both conditions are met, the heated surface area corresponding to the current temperature cluster is marked as a suspected coking area. The multi-point pressure data is converted into a frequency domain signal and subjected to Fourier transform to obtain a spectral distribution. The low-frequency components of the spectral distribution are extracted and the main frequency components are retained. The main frequency components are then subjected to inverse Fourier transform to reconstruct the pressure reference curve.

3. The method according to claim 1, characterized in that, By combining pre-acquired measured pressure distribution data to determine the pressure deviation vector, and by combining the suspected coking area to obtain quantitative indicators of coking degree, including: Obtain the measured pressure distribution data at the time corresponding to the pressure reference curve, align the pressure reference curve and the measured pressure distribution data according to spatial coordinates, calculate the difference between the pressure value of the pressure reference curve at each spatial coordinate point and the pressure value of the measured pressure distribution data at the corresponding spatial coordinate point, and arrange them according to the spatial coordinate order to obtain the pressure deviation vector. Based on the spatial coordinates of the measured pressure distribution data, a spatial grid of the heated surface is constructed. The pressure deviation vector is mapped onto the spatial grid of the heated surface according to the spatial coordinates to form a pressure deviation distribution field. The spatial coordinate range of the suspected coking area is extracted. The pressure deviation values ​​of the grid nodes corresponding to the spatial coordinate range are extracted from the pressure deviation distribution field, and spatial gradient calculation is performed to obtain the pressure deviation gradient vector. Principal component analysis is performed on the pressure deviation gradient vector to obtain the dominant gradient direction and the variance contribution rate corresponding to each principal component. The angle between the dominant gradient direction and the flue gas flow direction is calculated to obtain the flow direction deviation. The variance contribution rate is used as a weight to perform a weighted summation of the pressure deviation values ​​of each grid node in the suspected coking area. The coking degree quantification index is obtained by combining the flow direction deviation and the spatial range of the suspected coking area.

4. The method according to claim 1, characterized in that, Historical boiler combustion conditions are obtained and burner injection angle parameters are extracted. Heat load intensity distribution is calculated using a ray tracing algorithm, and a coking risk assessment matrix is ​​constructed by combining the aforementioned coking degree quantification index. This includes: Historical boiler combustion conditions are obtained from the historical database, and the injection angle parameters and injection speed parameters of each burner are extracted from the historical boiler combustion conditions. Taking the spatial position of the burner as the starting point of the ray, the injection angle parameter determines the propagation direction of the ray, the injection speed parameter determines the propagation step size of the ray, and the ray is gradually tracked along the propagation direction and the intersection position of the ray with each spatial coordinate point of the heated surface is recorded. The cumulative number of times each spatial coordinate point is passed by the ray is counted and the heat load intensity value is determined. The heat load intensity values ​​of all spatial coordinate points are arranged according to the spatial coordinates to obtain the heat load intensity distribution. The spatial coordinate range of the suspected coking area is obtained. The heat load intensity value corresponding to the spatial coordinate range is extracted from the heat load intensity distribution and matched with the coking degree quantification index according to the spatial coordinates to obtain a paired dataset. The heat load intensity value in the paired dataset is divided into intervals to obtain the heat load intensity level. The coking degree quantification index in the paired dataset is divided into intervals to obtain the coking degree level. The heat load intensity level is used as the row index and the coking degree level is used as the column index. The coking risk assessment matrix is ​​constructed by traversing each spatial coordinate point in the paired dataset.

5. The method according to claim 1, characterized in that, The coking risk assessment matrix is ​​topologically sorted and a directed weighted graph is constructed using pre-acquired flue gas flow paths. Based on the directed weighted graph, the coking propagation coefficient is calculated, and coking acceleration nodes are determined using Markov chains. Based on the coking acceleration nodes, a sootblower activation sequence is determined, and a time interval allocation scheme is generated using the coking propagation coefficient, including: The elements in the coking risk assessment matrix are sorted in descending order according to their numerical values ​​to obtain an element sequence. The dependencies between different matrix elements are determined based on the heat load intensity level and coking degree level corresponding to each element, and the element sequence is topologically sorted to obtain a priority sequence. Project the spatial coordinate points corresponding to the elements in the priority sequence along the flue gas flow path to obtain the projection position. Use the spatial coordinate points as graph nodes, determine the adjacent relationship between the graph nodes according to the projection position as the edge connection relationship, and construct a directed weighted graph using the elements of each spatial coordinate point in the coking risk assessment matrix as the node weight value. The outgoing edge propagation intensity is obtained by traversing the directed weighted graph and calculating the ratio of the weight value corresponding to the current graph node to the weight value corresponding to the downstream node. The outgoing edge propagation intensity of each graph node is summed to obtain the coking propagation coefficient. A Markov chain is constructed with the graph nodes as states and the coking propagation coefficient as the state transition probability, and a steady-state probability distribution is obtained by iterative calculation. Based on the steady-state probability distribution, coking acceleration nodes are determined. The spatial coordinates of each coking acceleration node are extracted and the corresponding sootblower number is queried. The sootblower start sequence is determined according to the order of the coking acceleration nodes on the flue gas flow path and the sootblower number. The time interval is determined by combining the coking propagation coefficient between adjacent coking acceleration nodes and a time interval allocation scheme is generated.

6. The method according to claim 1, characterized in that, Acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the sootblower start-up sequence, calculate the duration of operation, including: The steam pipeline network monitoring system acquires steam pipeline network pressure fluctuation data within a preset time period and performs time-domain decomposition to obtain a pressure reference component and a pressure fluctuation component. Spectral analysis is performed on the pressure fluctuation component to obtain the dominant frequency component and its corresponding amplitude. Based on the amplitude corresponding to the dominant frequency component, the pressure fluctuation intensity is calculated using an energy accumulation method. The steady-state available steam flow rate is calculated based on the pressure reference component. The flow loss corresponding to the safety margin is calculated based on the pressure fluctuation intensity to obtain the fluctuation margin. Based on the steady-state available steam flow rate and the fluctuation margin, the upper limit of the available steam flow rate is obtained using a margin deduction method. Extract the number of each sootblower in the sootblower start-up sequence and query the corresponding rated steam consumption and rated operating time. Obtain the coking propagation coefficient of the coking acceleration node corresponding to each sootblower and perform normalization processing to obtain the coking propagation weight corresponding to each sootblower. Based on the rated steam consumption and the coking propagation weight, obtain the weighted steam consumption by weighted summation. Determine the sufficiency of the available steam flow rate by comparing the upper limit of available steam flow rate with the weighted steam consumption to obtain the determination result. Based on the determination result, combine the rated operating time and the coking propagation weight to solve for the operating duration of each sootblower.

7. The method according to claim 1, characterized in that, Based on the duration of the action and the time interval allocation scheme, the total execution time is determined and it is judged whether it exceeds the time window constraint. If it does, the duration is compressed proportionally to obtain a corrected duration, and combined with the soot blower start sequence, a soot blowing control command sequence is generated and issued for execution, including: The time interval between adjacent sootblowers in the sootblower start-up sequence is extracted from the time interval allocation scheme and accumulated with the duration of action to obtain the total execution time. The current boiler operating condition data is obtained and the current load status and flue gas temperature are extracted. The corresponding time window constraint upper limit is queried according to the current load status and the flue gas temperature. It is determined whether the total execution time is greater than the time window constraint upper limit. If it is greater, the ratio coefficient between the total execution time and the time window constraint upper limit is calculated. The compression operation is performed based on the duration of action and the ratio coefficient to obtain the corrected duration. The steady-state probability value of the coking acceleration node corresponding to each sootblower is extracted and the priority weight is calculated. Based on the priority weight, the correction duration is redistributed and adjusted to obtain the optimal duration. Based on the optimal duration and the sootblower number in the sootblower start sequence, the start time and stop time of each sootblower are generated sequentially and encapsulated to obtain sootblowing control instructions. The sootblowing control instructions are arranged in the order of start time to generate a sootblowing control instruction sequence and sent to the sootblower control unit for execution through the industrial control network.

8. A soot blowing control optimization system for preventing coking in thermal power unit boilers, used to implement the method described in any one of claims 1-7, characterized in that, include: The coking sensing module is used to collect boiler heating surface temperature data and flue gas flow field pressure distribution data. It performs spatiotemporal clustering analysis on the heating surface temperature data to identify suspected coking areas, performs Fourier transform on the flue gas flow field pressure distribution data to reconstruct the pressure reference curve, determines the pressure deviation vector by combining the pre-acquired measured pressure distribution data, and obtains the coking degree quantification index by combining the suspected coking areas. The risk assessment module is used to obtain historical boiler combustion conditions and extract burner injection angle parameters, calculate heat load intensity distribution through ray tracing algorithm, and construct a coking risk assessment matrix by combining the coking degree quantification index. The sequence generation module is used to perform topological sorting on the coking risk assessment matrix and construct a directed weighted graph in combination with the pre-acquired flue gas flow path. Based on the directed weighted graph, the coking propagation coefficient is calculated and the coking acceleration node is determined through a Markov chain. Based on the coking acceleration node, the soot blower start sequence is determined and a time interval allocation scheme is generated in combination with the coking propagation coefficient. The soot blowing control module is used to acquire steam network pressure fluctuation data and calculate the upper limit of available steam flow. Based on the upper limit of available steam flow and the soot blower start sequence, it calculates the duration of action. Based on the duration of action and the time interval allocation scheme, it determines the total execution time and determines whether it exceeds the time window constraint. If it exceeds the constraint, it compresses the duration proportionally to obtain a corrected duration and combines it with the soot blower start sequence to generate a soot blowing control command sequence for execution.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images