Method and device for detecting coking in furnace

Abstract

The invention discloses a method and a device for detecting coking in a furnace. The method comprises the following steps: generating a coking evaluation value according to historical coking influence parameters of each detection point; setting an important grade and a monitoring time node of each detection point according to the coking evaluation value, and obtaining a real-time detection image and real-time related data of the detection point; generating a coking risk coefficient of the current detection point according to the real-time related data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generating a predicted coking reason of the current detection point; a preset coking image library is screened out according to the predicted coking reasons, and the preset coking amount of the preset coking image with the maximum similarity is set as the predicted coking amount of the real-time detection image; whether an alarm instruction is generated or not is judged according to the predicted coking reason and the predicted coking amount of the detection point and the importance level of the corresponding detection point, the judgment accuracy of the coking position and the coking amount in the furnace is improved, and safe operation of a unit is guaranteed.

Description

Technical Field

[0001] The present application relates to the technical field of in-furnace coking detection, and in particular to a method and device for detecting in-furnace coking. Background Art

[0002] In recent years, affected by the high coal prices, various thermal power plants have been in a loss-making or low-profit state for a long time. They generally adopt the method of blending and burning a large amount of low-priced economic coal to increase profits. Among them, blending and burning low-ash melting point coal is an important measure. When blending and burning low-ash melting point and high-alkali Xinjiang coal, the presence of alkali metals will cause coking problems in other types of coal blended into the furnace.

[0003] In the prior art, the method of judging whether coking occurs only by the changes in the smoke temperature and wall temperature in the furnace is used. It is difficult to accurately judge the position and amount of coking in the furnace, which endangers the safe operation of the unit and causes huge economic losses. Therefore, there is an urgent need for a method and device for detecting coking in the furnace to improve the accuracy of judging the position and amount of coking in the furnace and ensure the safe operation of the unit. Summary of the invention

[0004] In order to solve the above-mentioned technical problems, the present application provides a method and device for detecting coking in a furnace. By setting a monitoring time node for each detection point, real-time detection images and real-time related data of each detection point are obtained according to the monitoring time node, and a predicted coking cause is generated according to the real-time related data, and a predicted coking amount is generated according to the real-time detection image. According to the predicted coking cause, predicted coking amount and importance level of each detection point, it is determined whether to send an alarm signal, thereby improving the accuracy of determining the coking position and coking amount in the furnace and ensuring safe operation of the unit.

[0005] In some embodiments of the present application, a method for detecting coking in a furnace is provided, comprising:

[0006] Preset multiple detection points and generate a coking evaluation value based on the historical coking influence parameters of each detection point;

[0007] According to the coking evaluation value, the importance level and monitoring time node of each detection point are set, and the real-time detection image and real-time related data of the corresponding detection point are obtained based on the detection device and the monitoring time node;

[0008] Generate a coking risk coefficient for the current detection point based on real-time related data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generate a predicted coking cause for the current detection point;

[0009] According to the predicted coking cause, a preset coking image library corresponding to the detection point is selected, and the similarity between the preset coking image in the preset coking image library and the real-time detection image is calculated, and the preset coking amount of the preset coking image with the greatest similarity is set as the predicted coking amount of the real-time detection image;

[0010] Whether to generate an alarm instruction is determined based on the predicted coking cause, predicted coking amount and importance level of the corresponding detection point.

[0011] In some embodiments of the present application, a coking evaluation value is generated according to the historical coking influence parameter of each detection point, including:

[0012] Establish multiple coking impact evaluation indicators;

[0013] Obtain multiple historical coking logs of each detection point, and extract historical coking influence parameters in each historical coking log;

[0014] Determine the associated coking influence parameter of each coking influence evaluation index according to the degree of association between the historical coking influence parameter and the coking influence evaluation index in the same historical coking log, and generate a reference evaluation value of the corresponding coking influence evaluation index according to the associated coking influence parameter;

[0015] Generate an initial coking evaluation value of the corresponding historical coking log according to the reference evaluation values ​​of all coking impact evaluation indicators in the same historical coking log;

[0016] The calculation formula of the initial coking evaluation value is:

[0017]

[0018] Wherein, J is the initial coking evaluation value, n is the total number of coking impact evaluation indicators, bi is the reference evaluation value of the i-th coking impact evaluation indicator, and αi is the weight coefficient of the i-th coking impact evaluation indicator;

[0019] The initial coking evaluation values ​​of all historical coking logs of each detection point are averaged to obtain a coking evaluation value, and the importance level and monitoring time node of the corresponding detection point are set according to the coking evaluation mean.

[0020] In some embodiments of the present application, the importance level and monitoring time node of each detection point are set according to the coking evaluation value, including:

[0021] Presetting a first preset coking evaluation value threshold and a second preset coking evaluation value threshold;

[0022] When the coking evaluation value of the detection point is less than the first preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the first preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the third preset monitoring time interval t3;

[0023] When the coking evaluation value of the detection point is between the first preset coking evaluation value threshold and the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the second preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the second preset monitoring time interval t2;

[0024] When the coking evaluation value of the detection point is greater than the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the third preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the first preset monitoring time interval t1.

[0025] In some embodiments of the present application, the coking risk coefficient of the current detection point is generated according to the real-time related data, including:

[0026] Acquire real-time relevant data, compare the real-time relevant data with standard relevant data, and obtain data deviation values ​​between the real-time relevant data and the standard relevant data;

[0027] Obtaining a preset data deviation value interval of each real-time relevant data of each detection point, dividing the preset data deviation value interval of each real-time relevant data of each detection point into a plurality of preset data deviation value sub-intervals based on a coking influence factor and an operation stability factor, and configuring a corresponding preset coking sub-risk coefficient for each preset data deviation value sub-interval;

[0028] According to the correspondence between the data deviation value of each real-time relevant data and the standard relevant data of the current detection point and the preset data deviation value sub-interval, the preset coking sub-risk coefficient of the preset data deviation value sub-interval in which the data deviation value is located is set as the coking sub-risk coefficient of the corresponding real-time relevant data;

[0029] The coking risk coefficient is generated by combining the coking sub-risk coefficients of all real-time related data of the current detection point and the weight coefficients of the corresponding real-time related data.

[0030] In some embodiments of the present application, if the coking risk factor is greater than a preset risk factor threshold, a predicted coking cause of the current detection point is generated, including:

[0031] If the coking risk coefficient of the detection point is greater than the preset risk coefficient threshold, the coking risk sub-coefficient of each real-time relevant data of the corresponding detection point is compared with the preset risk sub-coefficient threshold, and the real-time relevant data with the coking risk sub-coefficient greater than the preset risk sub-coefficient threshold is screened out and set as abnormal relevant data;

[0032] Based on the similarity between the real-time operating condition and the historical operating condition of the historical coking log of the corresponding detection point, the historical coking logs having a similarity greater than a preset similarity threshold are screened out, and the historical abnormality-related data of the screened historical coking logs are determined;

[0033] Compare all abnormality-related data of the same detection point with the historical abnormality-related data of the screened historical coking logs to obtain a repetition ratio, and remove the historical coking logs with a repetition ratio less than a preset repetition ratio threshold and the historical abnormality-related data in the historical coking logs;

[0034] Obtain the historical coking causes and historical abnormality-related data in the remaining historical coking logs;

[0035] If all historical coking causes are the same, the current historical coking cause is set as the predicted coking cause. If the historical coking causes are different, a historical data change curve for each historical abnormality-related data in the remaining historical coking logs in the corresponding historical monitoring period is generated;

[0036] Generate a data change curve of the abnormal related data in the current monitoring period, and obtain the change trend of the abnormal related data and the change value at each monitoring time node according to the data change curve;

[0037] Compare the change trend of the abnormality-related data with the historical change trend of the corresponding historical abnormality-related data in the remaining historical coking logs in the historical data change curve to obtain the change trend similarity;

[0038] If the change trend similarity is greater than the preset change trend similarity threshold, then a similar curve segment of the change trend of the abnormality-related data is intercepted, and the corresponding historical abnormality-related data in the similar curve segment is obtained according to the monitoring time interval between the monitoring time nodes, and the historical change value of the historical abnormality-related data at the corresponding monitoring time node is obtained;

[0039] Generate a change value similarity based on the change value difference of the abnormal related data at each monitoring time node and the historical change value of the corresponding historical abnormal related data;

[0040] Generate change similarity between abnormal related data and corresponding historical abnormal related data based on change trend similarity and change value similarity;

[0041] The calculation formula of the change similarity is:

[0042]

[0043] Wherein, B is the change similarity, b1 is the change trend similarity, l1 is the weight coefficient of the change trend similarity, b2 is the change value similarity conversion coefficient, u is the number of monitoring time nodes, ΔY is the selection coefficient, if (b1-b')>0, ΔY=1; if (b1-b')<0, ΔY=0, b' is the preset change trend similarity threshold, do is the change value at the oth monitoring time node, do' is the historical change value at the oth monitoring time node;

[0044] Generate a comprehensive change similarity based on the repetition ratio and change similarity of the historical anomaly-related data and the current anomaly-related data in the same remaining historical coking log;

[0045]

[0046] Among them, B′ is the comprehensive change similarity, v1 is the number of historical abnormal related data that appear in the remaining historical coking logs, v0 is the total number of abnormal related data of the corresponding detection point, Bc is the change similarity between the cth abnormal related data and the corresponding historical abnormal related data, and Qc is the weight coefficient of the cth abnormal related data;

[0047] The historical coking causes of the remaining historical coking logs with the largest comprehensive change similarity are set as the predicted coking causes of the corresponding detection points.

[0048] In some embodiments of the present application, the method further comprises:

[0049] If the coking risk factor is less than the preset risk factor threshold, the risk factor difference between the coking risk factor and the preset risk factor threshold is calculated, and the monitoring time interval of the remaining monitoring time nodes of the current monitoring cycle of the corresponding detection point is corrected according to the risk factor difference;

[0050] Presetting a first preset risk factor difference interval, a second preset risk factor difference interval, a third preset risk factor difference interval and a fourth preset risk factor difference interval;

[0051] When the risk coefficient difference is within the first preset risk coefficient difference interval, a first preset correction coefficient r1 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r1*ti;

[0052] When the risk coefficient difference is within the second preset risk coefficient difference interval, a second preset correction coefficient r2 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r2*ti;

[0053] When the risk coefficient difference is within the third preset risk coefficient difference interval, the third preset correction coefficient r3 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r3*ti;

[0054] When the risk coefficient difference is in the fourth preset risk coefficient difference interval, the fourth preset correction coefficient r4 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r4*ti, where i=1, 2, 3.

[0055] In some embodiments of the present application, a preset coking image library corresponding to a detection point is screened out according to the predicted coking cause, including:

[0056] Filtering a preset focusing image library corresponding to the detection point according to the predicted focusing cause of the detection point, wherein the preset focusing image library includes a plurality of preset focusing images corresponding to the detection point under the current predicted focusing cause, and a preset focusing image is associated with a corresponding preset focusing amount;

[0057] A similarity analysis is performed between the real-time detection image and each preset focus image in the preset focus image library, and the preset focus amount of the preset focus image with a similarity greater than a preset similarity threshold and the greatest similarity is set as the predicted focus amount of the real-time detection image.

[0058] In some embodiments of the present application, calculating the similarity between a preset focus image in a preset focus image library and a real-time detection image includes:

[0059] Performing image processing on the real-time detection image and each preset focus image to obtain a real-time grayscale image of the real-time detection image and a preset grayscale focus image of each preset focus image, wherein the real-time grayscale image includes a plurality of local sub-regions, and the preset grayscale focus image includes a plurality of local focus sub-regions;

[0060] Determine a plurality of pixels in each local sub-region of the real-time grayscale image, extract a first pixel feature value of each pixel, and extract a second pixel feature value of a corresponding pixel in a corresponding local focus sub-region of each preset grayscale focus image based on the same region position and pixel position;

[0061] Calculating the pixel feature difference between each first pixel feature value of the real-time grayscale image and the corresponding second pixel feature value of the same preset grayscale focus image, performing difference analysis on the pixel feature differences, and obtaining the difference degree of each pixel feature difference;

[0062] Subtract the difference degree of each pixel feature difference from the preset difference degree threshold to obtain the difference degree difference value;

[0063] Generate the similarity between the real-time grayscale image and the corresponding preset grayscale focus image according to the difference value of the difference degree of the difference degree of the multiple pixels;

[0064] The calculation formula of the similarity is:

[0065]

[0066] Among them, X is the similarity, x0 is the similarity conversion coefficient, k1 is the weight coefficient of the first local sub-region in the real-time grayscale image, m1 is the number of pixels in the first local sub-region, F1 s1 is the difference between the pixel feature difference between the s1th pixel point and the corresponding pixel point in the first local sub-region, w0 is the preset difference degree threshold, f1 s1 is the weight coefficient of the s1th pixel in the first local sub-region, k2 is the weight coefficient of the second local sub-region in the real-time grayscale image, m2 is the number of pixels in the second local sub-region, F2 s2 is the difference between the pixel feature difference between the s2th pixel and the corresponding pixel in the second local sub-region, f2 s2 is the weight coefficient of the s2th pixel in the 2nd local subregion, kg is the weight coefficient of the gth local subregion in the real-time grayscale image, mg is the number of pixels in the gth local subregion, Fg sg is the difference between the pixel feature difference between the sgth pixel in the gth local sub-region and the corresponding pixel, fg sg is the weight coefficient of the sg-th pixel in the g-th local sub-region.

[0067] In some embodiments of the present application, judging whether to generate an alarm instruction according to the predicted coking cause of the detection point, the predicted coking amount and the importance level of the corresponding detection point includes:

[0068] Generate a predicted coking evaluation value of the current detection point according to the predicted coking cause and predicted coking amount of the current detection point;

[0069] Set the coking amount threshold under different coking causes according to the importance level of the current detection point;

[0070] The predicted coking amount of the predicted coking cause is compared with the coking amount threshold set under the corresponding coking cause. If the predicted coking amount is less than the coking amount threshold, no alarm signal is sent. If the predicted coking amount is greater than the coking amount threshold, an alarm signal is sent.

[0071] In some embodiments of the present application, a device for detecting coking in a furnace is also included:

[0072] A setting module, used to pre-set a plurality of detection points and generate a coking evaluation value according to the historical coking influence parameters of each detection point;

[0073] An acquisition module is used to set the importance level and monitoring time node of each detection point according to the coking evaluation value, and acquire the real-time detection image and real-time related data of the corresponding detection point based on the detection device and the monitoring time node;

[0074] A generation module, used to generate a coking risk coefficient of the current detection point according to real-time relevant data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generate a predicted coking cause of the current detection point;

[0075] A prediction module is used to select a preset focusing image library corresponding to the detection point according to the predicted focusing cause, calculate the similarity between the preset focusing image in the preset focusing image library and the real-time detection image, and set the preset focusing amount of the preset focusing image with the greatest similarity as the predicted focusing amount of the real-time detection image;

[0076] The alarm module is used to determine whether to generate an alarm instruction based on the predicted coking cause and predicted coking amount of the detection point and the importance level of the corresponding detection point.

[0077] Compared with the prior art, the method and device for detecting coking in a furnace in the embodiment of the present application have the following beneficial effects:

[0078] By setting the monitoring time node for each detection point, obtaining the real-time detection image and real-time related data of each detection point according to the monitoring time node, generating the predicted coking cause based on the real-time related data, generating the predicted coking amount based on the real-time detection image, and judging whether to send an alarm signal based on the predicted coking cause, predicted coking amount and importance level of each detection point, the accuracy of judging the coking position and coking amount in the furnace is improved to ensure the safe operation of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a schematic flow chart of a method for detecting coking in a furnace in an embodiment of the present application;

[0080] Figure 2 It is a schematic diagram of a detection device for coking in a furnace in an embodiment of the present application. DETAILED DESCRIPTION

[0081] The specific implementation methods of the present application are further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application but are not intended to limit the scope of the present application.

[0082] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0083] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0084] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0085] like Figure 1 As shown, a method for detecting coking in a furnace according to an embodiment of the present application includes:

[0086] Step S101: pre-set a plurality of detection points, and generate a coking evaluation value according to the historical coking influence parameter of each detection point;

[0087] Step S102: setting the importance level and monitoring time node of each detection point according to the coking evaluation value, and acquiring the real-time detection image and real-time related data of the corresponding detection point based on the detection device and the monitoring time node;

[0088] Step S103: generating a coking risk coefficient of the current detection point according to the real-time related data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generating a predicted coking cause of the current detection point;

[0089] Step S104: Filter out a preset focusing image library corresponding to the detection point according to the predicted focusing cause, calculate the similarity between the preset focusing image in the preset focusing image library and the real-time detection image, and set the preset focusing amount of the preset focusing image with the greatest similarity as the predicted focusing amount of the real-time detection image;

[0090] Step S105: Determine whether to generate an alarm instruction based on the predicted coking cause and predicted coking amount of the detection point and the importance level of the corresponding detection point.

[0091] In this embodiment, the detection device is a full-view high-temperature camera robotic arm. Through wind and water combined cooling protection, the robotic arm extends into the furnace and carries a smart camera for full-view video monitoring to obtain high-definition images of each detection point. At the same time, it collects the real-time operating parameters of the boiler, filters out the real-time operating data on the impact of coking on different detection points, and sets them as real-time related data.

[0092] In this embodiment, the predicted coking cause of each detection point is determined by analyzing the real-time relevant data, and the real-time detection image is analyzed to determine the predicted coking amount of the corresponding detection point under the predicted coking cause. According to the predicted coking cause, the preset coking amount and the importance level of the detection point, it is determined whether to generate an alarm instruction, thereby improving the accuracy of the judgment of coking in the furnace, ensuring the safe and stable operation of the unit, and avoiding unnecessary economic losses.

[0093] In some embodiments of the present application, a coking evaluation value is generated according to the historical coking influence parameter of each detection point, including:

[0094] Establish multiple coking impact evaluation indicators;

[0095] Obtain multiple historical coking logs of each detection point, and extract historical coking influence parameters in each historical coking log;

[0096] Determine the associated coking influence parameter of each coking influence evaluation index according to the degree of association between the historical coking influence parameter and the coking influence evaluation index in the same historical coking log, and generate a reference evaluation value of the corresponding coking influence evaluation index according to the associated coking influence parameter;

[0097] Generate an initial coking evaluation value of the corresponding historical coking log according to the reference evaluation values ​​of all coking impact evaluation indicators in the same historical coking log;

[0098] The calculation formula of the initial coking evaluation value is:

[0099]

[0100] Wherein, J is the initial coking evaluation value, n is the total number of coking impact evaluation indicators, bi is the reference evaluation value of the i-th coking impact evaluation indicator, and αi is the weight coefficient of the i-th coking impact evaluation indicator;

[0101] The initial coking evaluation values ​​of all historical coking logs of each detection point are averaged to obtain a coking evaluation value, and the importance level and monitoring time node of the corresponding detection point are set according to the coking evaluation mean.

[0102] In this embodiment, the coking impact evaluation index is an index used to evaluate the impact of the coking phenomenon on the unit operating status, unit operating cost, unit operating efficiency, and unit maintenance cost.

[0103] In this embodiment, the initial coking evaluation value in each historical coking log is calculated to obtain the coking evaluation value of all historical coking logs of each detection point. The importance level of the detection point is set according to the coking evaluation value, thereby setting the corresponding monitoring time node to ensure the detection accuracy of detection points of different importance levels, timely discover the coking amount of the detection point and formulate response measures to avoid unit shutdown and huge economic losses.

[0104] In some embodiments of the present application, the importance level and monitoring time node of each detection point are set according to the coking evaluation value, including:

[0105] Presetting a first preset coking evaluation value threshold and a second preset coking evaluation value threshold;

[0106] When the coking evaluation value of the detection point is less than the first preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the first preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the third preset monitoring time interval t3;

[0107] When the coking evaluation value of the detection point is between the first preset coking evaluation value threshold and the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the second preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the second preset monitoring time interval t2;

[0108] When the coking evaluation value of the detection point is greater than the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the third preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the first preset monitoring time interval t1.

[0109] In this embodiment, the larger the coking evaluation value is, the greater the impact of coking at the corresponding detection point on the operating status, operating efficiency, and maintenance cost of the unit. Therefore, the monitoring time interval between the monitoring time nodes of the corresponding detection point should be smaller, so as to timely discover whether the detection point is coked and predict the coking amount, improve the detection efficiency and detection accuracy, ensure the stable operation of the unit and improve profitability.

[0110] In some embodiments of the present application, the coking risk coefficient of the current detection point is generated according to the real-time related data, including:

[0111] Acquire real-time relevant data, compare the real-time relevant data with standard relevant data, and obtain data deviation values ​​between the real-time relevant data and the standard relevant data;

[0112] Obtaining a preset data deviation value interval of each real-time relevant data of each detection point, dividing the preset data deviation value interval of each real-time relevant data of each detection point into a plurality of preset data deviation value sub-intervals based on a coking influence factor and an operation stability factor, and configuring a corresponding preset coking sub-risk coefficient for each preset data deviation value sub-interval;

[0113] According to the correspondence between the data deviation value of each real-time relevant data and the standard relevant data of the current detection point and the preset data deviation value sub-interval, the preset coking sub-risk coefficient of the preset data deviation value sub-interval in which the data deviation value is located is set as the coking sub-risk coefficient of the corresponding real-time relevant data;

[0114] The coking risk coefficient is generated by combining the coking sub-risk coefficients of all real-time related data of the current detection point and the weight coefficients of the corresponding real-time related data.

[0115] In this embodiment, the coking influence factor refers to the influence of multiple historical data deviation values ​​of each real-time related data on the coking degree of the current detection point, and the operating stability factor refers to the influence of coking phenomenon at the current detection point and different coking degrees on the stability of the unit's operating state. The preset coking sub-risk coefficient is calculated based on the influence of the historical data deviation values ​​of each preset data deviation value sub-interval on the historical coking degree and the influence of the historical coking degree on the stability of the unit's operating state.

[0116] In this embodiment, by calculating the coking risk coefficient, it is determined whether there is a coking risk at each detection point. If so, the cause of coking and the amount of coking are predicted, and the alarm signal is sent and adjusted in time to ensure stable operation of the unit. If not, the monitoring time node is adjusted to accurately obtain the real-time detection image of each detection point, improve the detection accuracy of each detection point, and avoid the situation where the coking phenomenon is not discovered in time.

[0117] In some embodiments of the present application, if the coking risk factor is greater than a preset risk factor threshold, a predicted coking cause of the current detection point is generated, including:

[0118] If the coking risk coefficient of the detection point is greater than the preset risk coefficient threshold, the coking risk sub-coefficient of each real-time relevant data of the corresponding detection point is compared with the preset risk sub-coefficient threshold, and the real-time relevant data with the coking risk sub-coefficient greater than the preset risk sub-coefficient threshold is screened out and set as abnormal relevant data;

[0119] Based on the similarity between the real-time operating condition and the historical operating condition of the historical coking log of the corresponding detection point, the historical coking logs having a similarity greater than a preset similarity threshold are screened out, and the historical abnormality-related data of the screened historical coking logs are determined;

[0120] Compare all abnormality-related data of the same detection point with the historical abnormality-related data of the screened historical coking logs to obtain a repetition ratio, and remove the historical coking logs with a repetition ratio less than a preset repetition ratio threshold and the historical abnormality-related data in the historical coking logs;

[0121] Obtain the historical coking causes and historical abnormality-related data in the remaining historical coking logs;

[0122] If all historical coking causes are the same, the current historical coking cause is set as the predicted coking cause. If the historical coking causes are different, a historical data change curve for each historical abnormality-related data in the remaining historical coking logs in the corresponding historical monitoring period is generated;

[0123] Generate a data change curve of the abnormal related data in the current monitoring period, and obtain the change trend of the abnormal related data and the change value at each monitoring time node according to the data change curve;

[0124] Compare the change trend of the abnormality-related data with the historical change trend of the corresponding historical abnormality-related data in the remaining historical coking logs in the historical data change curve to obtain the change trend similarity;

[0125] If the change trend similarity is greater than the preset change trend similarity threshold, then a similar curve segment of the change trend of the abnormality-related data is intercepted, and the corresponding historical abnormality-related data in the similar curve segment is obtained according to the monitoring time interval between the monitoring time nodes, and the historical change value of the historical abnormality-related data at the corresponding monitoring time node is obtained;

[0126] Generate a change value similarity based on the change value difference of the abnormal related data at each monitoring time node and the historical change value of the corresponding historical abnormal related data;

[0127] Generate change similarity between abnormal related data and corresponding historical abnormal related data based on change trend similarity and change value similarity;

[0128] The calculation formula of the change similarity is:

[0129]

[0130] Wherein, B is the change similarity, b1 is the change trend similarity, l1 is the weight coefficient of the change trend similarity, b2 is the change value similarity conversion coefficient, u is the number of monitoring time nodes, ΔY is the selection coefficient, if (b1-b')>0, ΔY=1; if (b1-b')<0, ΔY=0, b' is the preset change trend similarity threshold, do is the change value at the oth monitoring time node, do' is the historical change value at the oth monitoring time node;

[0131] Generate a comprehensive change similarity based on the repetition ratio and change similarity of the historical anomaly-related data and the current anomaly-related data in the same remaining historical coking log;

[0132]

[0133] Among them, B′ is the comprehensive change similarity, v1 is the number of historical abnormal related data that appear in the remaining historical coking logs, v0 is the total number of abnormal related data of the corresponding detection point, Bc is the change similarity between the cth abnormal related data and the corresponding historical abnormal related data, and Qc is the weight coefficient of the cth abnormal related data;

[0134] The historical coking causes of the remaining historical coking logs with the largest comprehensive change similarity are set as the predicted coking causes of the corresponding detection points.

[0135] In this embodiment, the repetition ratio refers to the number of repetitions of all abnormality-related data of the same detection point and the historical abnormality-related data in the same screened historical coking log / the total number of all abnormality-related data, and v1 / v0 is the repetition ratio.

[0136] In this embodiment, the change trend similarity is obtained by comparing the slope of the data change curve with the slope of the historical data change curve. The similar curve segment refers to a curve segment in the historical data change curve whose change trend similarity with the data change curve is greater than the preset change trend similarity.

[0137] In this embodiment, by comparing the real-time operating conditions with the historical operating conditions in the historical coking log of the corresponding detection point, the corresponding historical coking log is screened out, and the repetition ratio of historical abnormality-related data and current abnormality-related data in the historical coking log is calculated. Some historical coking logs are eliminated according to the repetition ratio to reduce the amount of data analysis and processing. The comprehensive change similarity is calculated based on the change similarity of historical abnormality-related data and abnormality-related data in the remaining historical coking logs and the repetition ratio to obtain a historical coking log that is similar to the real-time operating conditions, abnormality-related data, and change characteristics of abnormality-related data of the current detection point, thereby determining the predicted cause of coking, improving the accuracy of the predicted cause of coking, and laying the foundation for subsequent predicted coking amount.

[0138] In some embodiments of the present application, the method further comprises:

[0139] If the coking risk factor is less than the preset risk factor threshold, the risk factor difference between the coking risk factor and the preset risk factor threshold is calculated, and the monitoring time interval of the remaining monitoring time nodes of the current monitoring cycle of the corresponding detection point is corrected according to the risk factor difference;

[0140] Presetting a first preset risk factor difference interval, a second preset risk factor difference interval, a third preset risk factor difference interval and a fourth preset risk factor difference interval;

[0141] When the risk coefficient difference is within the first preset risk coefficient difference interval, a first preset correction coefficient r1 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r1*ti;

[0142] When the risk coefficient difference is within the second preset risk coefficient difference interval, a second preset correction coefficient r2 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r2*ti;

[0143] When the risk coefficient difference is within the third preset risk coefficient difference interval, the third preset correction coefficient r3 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r3*ti;

[0144] When the risk coefficient difference is in the fourth preset risk coefficient difference interval, the fourth preset correction coefficient r4 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r4*ti, where i=1, 2, 3.

[0145] In this embodiment, the first preset risk coefficient difference interval < the second preset risk coefficient difference interval < the third preset risk coefficient difference interval < the fourth preset risk coefficient difference interval, 0.8 < r1 < r2 < 1 < r3 < r4 < 1.2.

[0146] In this embodiment, the risk coefficient difference = preset risk coefficient threshold - coking risk coefficient. When the risk coefficient difference is larger, the coking risk coefficient is smaller, that is, the probability of coking at the corresponding detection point is smaller, and the monitoring time interval of the remaining monitoring time nodes can be appropriately increased. When the risk coefficient difference is smaller, the coking risk coefficient is larger, that is, the probability of coking at the corresponding detection point is larger, and the monitoring time interval of the remaining monitoring time nodes can be appropriately reduced to avoid the problem of untimely monitoring, resulting in inaccurate detection of coking in the furnace, inability to accurately judge the coking position and coking amount, and reduced detection efficiency.

[0147] In some embodiments of the present application, a preset coking image library corresponding to a detection point is screened out according to the predicted coking cause, including:

[0148] Filtering a preset focusing image library corresponding to the detection point according to the predicted focusing cause of the detection point, wherein the preset focusing image library includes a plurality of preset focusing images corresponding to the detection point under the current predicted focusing cause, and a preset focusing image is associated with a corresponding preset focusing amount;

[0149] A similarity analysis is performed between the real-time detection image and each preset focus image in the preset focus image library, and the preset focus amount of the preset focus image with a similarity greater than a preset similarity threshold and the greatest similarity is set as the predicted focus amount of the real-time detection image.

[0150] In some embodiments of the present application, calculating the similarity between a preset focus image in a preset focus image library and a real-time detection image includes:

[0151] Performing image processing on the real-time detection image and each preset focus image to obtain a real-time grayscale image of the real-time detection image and a preset grayscale focus image of each preset focus image, wherein the real-time grayscale image includes a plurality of local sub-regions, and the preset grayscale focus image includes a plurality of local focus sub-regions;

[0152] Determine a plurality of pixels in each local sub-region of the real-time grayscale image, extract a first pixel feature value of each pixel, and extract a second pixel feature value of a corresponding pixel in a corresponding local focus sub-region of each preset grayscale focus image based on the same region position and pixel position;

[0153] Calculating the pixel feature difference between each first pixel feature value of the real-time grayscale image and the corresponding second pixel feature value of the same preset grayscale focus image, performing difference analysis on the pixel feature differences, and obtaining the difference degree of each pixel feature difference;

[0154] Subtract the difference degree of each pixel feature difference from the preset difference degree threshold to obtain the difference degree difference value;

[0155] Generate the similarity between the real-time grayscale image and the corresponding preset grayscale focus image according to the difference value of the difference degree of the difference degree of the multiple pixels;

[0156] The calculation formula of the similarity is:

[0157]

[0158] Among them, X is the similarity, x0 is the similarity conversion coefficient, k1 is the weight coefficient of the first local sub-region in the real-time grayscale image, m1 is the number of pixels in the first local sub-region, F1 s1is the difference between the pixel feature difference between the s1th pixel point and the corresponding pixel point in the first local sub-region, w0 is the preset difference degree threshold, f1 s1 is the weight coefficient of the s1th pixel in the first local sub-region, k2 is the weight coefficient of the second local sub-region in the real-time grayscale image, m2 is the number of pixels in the second local sub-region, F2 s2 is the difference between the pixel feature difference between the s2th pixel and the corresponding pixel in the second local sub-region, f2 s2 is the weight coefficient of the s2th pixel in the 2nd local subregion, kg is the weight coefficient of the gth local subregion in the real-time grayscale image, mg is the number of pixels in the gth local subregion, Fg sg is the difference between the pixel feature difference between the sgth pixel in the gth local sub-region and the corresponding pixel, fg sg is the weight coefficient of the sg-th pixel in the g-th local sub-region.

[0159] In this embodiment, the image processing includes eliminating noise, outliers and non-overlapping window strategies, dividing the real-time detection image and the preset focus image into multiple local areas, laying a foundation for the subsequent similarity analysis efficiency and accuracy.

[0160] In this embodiment, s1=1, 2...m1, s2=1, 2...m2, sg=1, 2...mg.

[0161] In this embodiment, the pixel feature values ​​include but are not limited to grayscale values, color values, texture features, etc. The first pixel feature value of each pixel point in the real-time grayscale image is compared with the second pixel feature value of the pixel point at the same position in the preset focus grayscale image to obtain the pixel feature difference between the first pixel feature value and the second pixel feature value, and the degree of difference is analyzed to obtain the degree of difference of each pixel feature difference. The similarity between the real-time grayscale image and the preset focus grayscale image is calculated according to the degree of difference of the pixel feature differences of multiple pixels in each local sub-area, that is, the similarity between the real-time detection image and the preset focus image, thereby improving the calculation accuracy of the similarity and laying a foundation for determining the predicted amount of coking.

[0162] In some embodiments of the present application, judging whether to generate an alarm instruction according to the predicted coking cause of the detection point, the predicted coking amount and the importance level of the corresponding detection point includes:

[0163] Generate a predicted coking evaluation value of the current detection point according to the predicted coking cause and predicted coking amount of the current detection point;

[0164] Set the coking amount threshold under different coking causes according to the importance level of the current detection point;

[0165] The predicted coking amount of the predicted coking cause is compared with the coking amount threshold set under the corresponding coking cause. If the predicted coking amount is less than the coking amount threshold, no alarm signal is sent. If the predicted coking amount is greater than the coking amount threshold, an alarm signal is sent.

[0166] In this embodiment, by setting the coking amount threshold (maximum coking amount) of detection points of different importance levels under different coking causes, it is ensured that the coking amount of the corresponding detection points does not affect the normal operation of the unit, reduces maintenance costs, and improves the profitability of the enterprise.

[0167] In some embodiments of the present application, a device for detecting coking in a furnace is also included:

[0168] A setting module, used to pre-set a plurality of detection points and generate a coking evaluation value according to the historical coking influence parameters of each detection point;

[0169] An acquisition module is used to set the importance level and monitoring time node of each detection point according to the coking evaluation value, and acquire the real-time detection image and real-time related data of the corresponding detection point based on the detection device and the monitoring time node;

[0170] A generation module, used to generate a coking risk coefficient of the current detection point according to real-time relevant data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generate a predicted coking cause of the current detection point;

[0171] A prediction module is used to select a preset focusing image library corresponding to the detection point according to the predicted focusing cause, calculate the similarity between the preset focusing image in the preset focusing image library and the real-time detection image, and set the preset focusing amount of the preset focusing image with the greatest similarity as the predicted focusing amount of the real-time detection image;

[0172] The alarm module is used to determine whether to generate an alarm instruction based on the predicted coking cause and predicted coking amount of the detection point and the importance level of the corresponding detection point.

[0173] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present application. These improvements and substitutions should also be regarded as the scope of protection of the present application.

Claims

1. A method for detecting coking in a furnace, characterized in that: include: Preset multiple detection points and generate a coking evaluation value based on the historical coking influence parameters of each detection point; According to the coking evaluation value, the importance level and monitoring time node of each detection point are set, and the real-time detection image and real-time related data of the corresponding detection point are obtained based on the detection device and the monitoring time node; Generate a coking risk coefficient for the current detection point based on real-time related data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generate a predicted coking cause for the current detection point; According to the predicted coking cause, a preset coking image library corresponding to the detection point is selected, and the similarity between the preset coking image in the preset coking image library and the real-time detection image is calculated, and the preset coking amount of the preset coking image with the greatest similarity is set as the predicted coking amount of the real-time detection image; Whether to generate an alarm instruction is determined based on the predicted coking cause, predicted coking amount and importance level of the corresponding detection point.

2. The method for detecting coking in a furnace according to claim 1, characterized in that: Generate a coking evaluation value based on the historical coking impact parameters of each detection point, including: Establish multiple coking impact evaluation indicators; Obtain multiple historical coking logs of each detection point, and extract historical coking influence parameters in each historical coking log; Determine the associated coking influence parameter of each coking influence evaluation index according to the degree of association between the historical coking influence parameter and the coking influence evaluation index in the same historical coking log, and generate a reference evaluation value of the corresponding coking influence evaluation index according to the associated coking influence parameter; Generate an initial coking evaluation value of the corresponding historical coking log according to the reference evaluation values ​​of all coking impact evaluation indicators in the same historical coking log; The calculation formula of the initial coking evaluation value is: Wherein, J is the initial coking evaluation value, n is the total number of coking impact evaluation indicators, bi is the reference evaluation value of the i-th coking impact evaluation indicator, and αi is the weight coefficient of the i-th coking impact evaluation indicator; The initial coking evaluation values ​​of all historical coking logs of each detection point are averaged to obtain a coking evaluation value, and the importance level and monitoring time node of the corresponding detection point are set according to the coking evaluation mean.

3. The method for detecting coking in a furnace according to claim 2, characterized in that: The importance level and monitoring time node of each detection point are set according to the coking evaluation value, including: Presetting a first preset coking evaluation value threshold and a second preset coking evaluation value threshold; When the coking evaluation value of the detection point is less than the first preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the first preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the third preset monitoring time interval t3; When the coking evaluation value of the detection point is between the first preset coking evaluation value threshold and the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the second preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the second preset monitoring time interval t2; When the coking evaluation value of the detection point is greater than the second preset coking evaluation value threshold, the importance level of the corresponding detection point is set to the third preset importance level, and the monitoring time interval between the monitoring time nodes of the corresponding detection point is set to the first preset monitoring time interval t1.

4. The method for detecting coking in a furnace according to claim 3, characterized in that: Generate the coking risk factor of the current detection point based on real-time relevant data, including: Acquire real-time relevant data, compare the real-time relevant data with standard relevant data, and obtain data deviation values ​​between the real-time relevant data and the standard relevant data; Obtaining a preset data deviation value interval of each real-time relevant data of each detection point, dividing the preset data deviation value interval of each real-time relevant data of each detection point into a plurality of preset data deviation value sub-intervals based on a coking influence factor and an operation stability factor, and configuring a corresponding preset coking sub-risk coefficient for each preset data deviation value sub-interval; According to the correspondence between the data deviation value of each real-time relevant data and the standard relevant data of the current detection point and the preset data deviation value sub-interval, the preset coking sub-risk coefficient of the preset data deviation value sub-interval in which the data deviation value is located is set as the coking sub-risk coefficient of the corresponding real-time relevant data; The coking risk coefficient is generated by combining the coking sub-risk coefficients of all real-time related data of the current detection point and the weight coefficients of the corresponding real-time related data.

5. The method for detecting coking in a furnace according to claim 4, characterized in that: If the coking risk factor is greater than the preset risk factor threshold, the predicted coking cause of the current detection point is generated, including: If the coking risk coefficient of the detection point is greater than the preset risk coefficient threshold, the coking risk sub-coefficient of each real-time relevant data of the corresponding detection point is compared with the preset risk sub-coefficient threshold, and the real-time relevant data with the coking risk sub-coefficient greater than the preset risk sub-coefficient threshold is screened out and set as abnormal relevant data; Based on the similarity between the real-time operating condition and the historical operating condition of the historical coking log of the corresponding detection point, the historical coking logs having a similarity greater than a preset similarity threshold are screened out, and the historical abnormality-related data of the screened historical coking logs are determined; Compare all abnormality-related data of the same detection point with the historical abnormality-related data of the screened historical coking logs to obtain a repetition ratio, and remove the historical coking logs with a repetition ratio less than a preset repetition ratio threshold and the historical abnormality-related data in the historical coking logs; Obtain the historical coking causes and historical abnormality-related data in the remaining historical coking logs; If all historical coking causes are the same, the current historical coking cause is set as the predicted coking cause. If the historical coking causes are different, a historical data change curve for each historical abnormality-related data in the remaining historical coking logs in the corresponding historical monitoring period is generated; Generate a data change curve of the abnormal related data in the current monitoring period, and obtain the change trend of the abnormal related data and the change value at each monitoring time node according to the data change curve; Compare the change trend of the abnormality-related data with the historical change trend of the corresponding historical abnormality-related data in the remaining historical coking logs in the historical data change curve to obtain the change trend similarity; If the change trend similarity is greater than the preset change trend similarity threshold, then a similar curve segment of the change trend of the abnormality-related data is intercepted, and the corresponding historical abnormality-related data in the similar curve segment is obtained according to the monitoring time interval between the monitoring time nodes, and the historical change value of the historical abnormality-related data at the corresponding monitoring time node is obtained; Generate a change value similarity based on the change value difference of the abnormal related data at each monitoring time node and the historical change value of the corresponding historical abnormal related data; Generate change similarity between abnormal related data and corresponding historical abnormal related data based on change trend similarity and change value similarity; The calculation formula of the change similarity is: Wherein, B is the change similarity, b1 is the change trend similarity, l1 is the weight coefficient of the change trend similarity, b2 is the change value similarity conversion coefficient, u is the number of monitoring time nodes, ΔY is the selection coefficient, if (b1-b')>0, ΔY=1; if (b1-b')<0, ΔY=0, b' is the preset change trend similarity threshold, do is the change value at the oth monitoring time node, do' is the historical change value at the oth monitoring time node; Generate a comprehensive change similarity based on the repetition ratio and change similarity of the historical anomaly-related data and the current anomaly-related data in the same remaining historical coking log; Among them, B′ is the comprehensive change similarity, v1 is the number of historical abnormal related data that appear in the remaining historical coking logs, v0 is the total number of abnormal related data of the corresponding detection point, Bc is the change similarity between the cth abnormal related data and the corresponding historical abnormal related data, and Qc is the weight coefficient of the cth abnormal related data; The historical coking causes of the remaining historical coking logs with the largest comprehensive change similarity are set as the predicted coking causes of the corresponding detection points.

6. The method for detecting coking in a furnace according to claim 5, characterized in that: Also includes: If the coking risk factor is less than the preset risk factor threshold, the risk factor difference between the coking risk factor and the preset risk factor threshold is calculated, and the monitoring time interval of the remaining monitoring time nodes of the current monitoring cycle of the corresponding detection point is corrected according to the risk factor difference; Presetting a first preset risk factor difference interval, a second preset risk factor difference interval, a third preset risk factor difference interval and a fourth preset risk factor difference interval; When the risk coefficient difference is within the first preset risk coefficient difference interval, a first preset correction coefficient r1 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r1*ti; When the risk coefficient difference is within the second preset risk coefficient difference interval, a second preset correction coefficient r2 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r2*ti; When the risk coefficient difference is within the third preset risk coefficient difference interval, the third preset correction coefficient r3 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r3*ti; When the risk coefficient difference is in the fourth preset risk coefficient difference interval, the fourth preset correction coefficient r4 is selected to correct the monitoring time interval of the corresponding detection point, and the corrected monitoring time interval is r4*ti, where i=1, 2, 3.

7. The method for detecting coking in a furnace according to claim 6, characterized in that: The preset coking image library of the corresponding detection point is selected according to the predicted coking cause, including: Filtering a preset focusing image library corresponding to the detection point according to the predicted focusing cause of the detection point, wherein the preset focusing image library includes a plurality of preset focusing images corresponding to the detection point under the current predicted focusing cause, and a preset focusing image is associated with a corresponding preset focusing amount; A similarity analysis is performed between the real-time detection image and each preset focus image in the preset focus image library, and the preset focus amount of the preset focus image with a similarity greater than a preset similarity threshold and the greatest similarity is set as the predicted focus amount of the real-time detection image.

8. The method for detecting coking in a furnace according to claim 7, characterized in that: Calculate the similarity between the preset defocused image in the preset defocused image library and the real-time detection image, including: Performing image processing on the real-time detection image and each preset focus image to obtain a real-time grayscale image of the real-time detection image and a preset grayscale focus image of each preset focus image, wherein the real-time grayscale image includes a plurality of local sub-regions, and the preset grayscale focus image includes a plurality of local focus sub-regions; Determine a plurality of pixels in each local sub-region of the real-time grayscale image, extract a first pixel feature value of each pixel, and extract a second pixel feature value of a corresponding pixel in a corresponding local focus sub-region of each preset grayscale focus image based on the same region position and pixel position; Calculating the pixel feature difference between each first pixel feature value of the real-time grayscale image and the corresponding second pixel feature value of the same preset grayscale focus image, performing difference analysis on the pixel feature difference, and obtaining the difference degree of each pixel feature difference; Subtract the difference degree of each pixel feature difference from the preset difference degree threshold to obtain the difference degree difference value; Generate the similarity between the real-time grayscale image and the corresponding preset grayscale focus image according to the difference value of the difference degree of the difference degree of the multiple pixels; The calculation formula of the similarity is: Among them, X is the similarity, x0 is the similarity conversion coefficient, k1 is the weight coefficient of the first local sub-region in the real-time grayscale image, m1 is the number of pixels in the first local sub-region, F1 s1 is the difference between the pixel feature difference between the s1th pixel point and the corresponding pixel point in the first local sub-region, w0 is the preset difference degree threshold, f1 s1 is the weight coefficient of the s1th pixel in the first local sub-region, k2 is the weight coefficient of the second local sub-region in the real-time grayscale image, m2 is the number of pixels in the second local sub-region, F2 s2 is the difference between the pixel feature difference between the s2th pixel and the corresponding pixel in the second local sub-region, f2 s2 is the weight coefficient of the s2th pixel in the 2nd local subregion, kg is the weight coefficient of the gth local subregion in the real-time grayscale image, mg is the number of pixels in the gth local subregion, Fg sg is the difference between the pixel feature difference between the sgth pixel in the gth local sub-region and the corresponding pixel, fg sg is the weight coefficient of the sg-th pixel in the g-th local sub-region.

9. The method for detecting coking in a furnace according to claim 8, characterized in that: Whether to generate an alarm instruction is determined based on the predicted coking cause, predicted coking amount and importance level of the corresponding detection point, including: Generate a predicted coking evaluation value of the current detection point according to the predicted coking cause and predicted coking amount of the current detection point; Set the coking amount threshold under different coking causes according to the importance level of the current detection point; The predicted coking amount of the predicted coking cause is compared with the coking amount threshold set under the corresponding coking cause. If the predicted coking amount is less than the coking amount threshold, no alarm signal is sent. If the predicted coking amount is greater than the coking amount threshold, an alarm signal is sent.

10. A method for detecting coking in a furnace, characterized in that: include: A setting module, used to pre-set a plurality of detection points and generate a coking evaluation value according to the historical coking influence parameters of each detection point; An acquisition module is used to set the importance level and monitoring time node of each detection point according to the coking evaluation value, and acquire the real-time detection image and real-time related data of the corresponding detection point based on the detection device and the monitoring time node; A generation module, used to generate a coking risk coefficient of the current detection point according to real-time relevant data, and if the coking risk coefficient is greater than a preset risk coefficient threshold, generate a predicted coking cause of the current detection point; A prediction module is used to select a preset focusing image library corresponding to the detection point according to the predicted focusing cause, calculate the similarity between the preset focusing image in the preset focusing image library and the real-time detection image, and set the preset focusing amount of the preset focusing image with the greatest similarity as the predicted focusing amount of the real-time detection image; The alarm module is used to determine whether to generate an alarm instruction based on the predicted coking cause and predicted coking amount of the detection point and the importance level of the corresponding detection point.

Images