Boiler fuel optimization control method and system based on flame center height monitoring

Abstract

The application provides a boiler fuel optimization control method and system based on flame center height monitoring, and relates to the technical field of automatic boiler control. The method comprises the following steps: at multiple time points in a monitoring period, a binocular camera arranged at multiple observation holes of a boiler is used to shoot flame images and infrared temperature measurement images in a furnace; a flame combustion stability index is determined; a flame shape index is determined; a flame combustion state index is determined; it is determined whether the control parameter of the burner needs to be adjusted; the adjustment parameter of each burner is determined; the adjusted fuel flow and the adjusted air flow of each burner are determined; and each burner is set. According to the application, the flame combustion state index can be determined based on the flame images and the infrared temperature measurement images, then it is determined whether the control parameter of the burner needs to be adjusted, and the control parameter of the burner in the next monitoring period is determined, so that the accuracy and real-time performance of fuel optimization are improved, and the fuel combustion efficiency of the boiler is improved.

Description

Technical Field

[0001] This invention relates to the field of boiler automatic control technology, and in particular to a boiler fuel optimization control method and system based on flame center height monitoring. Background Technology

[0002] Boilers are crucial energy conversion devices in industrial production. During boiler operation, it is essential to monitor the combustion state of the boiler flame in real time and optimize the fuel and air flow rates of the burners to reduce production costs, energy consumption, and pollutant emissions. Traditional boiler control methods typically monitor the combustion state of the boiler flame and optimize fuel and air flow rates based on parameters such as furnace temperature and flue gas oxygen content. However, this approach suffers from problems such as untimely monitoring and inaccurate adjustments. It is difficult to evaluate the flame combustion state in real time from multiple dimensions, including flame morphology and temperature, and it is also challenging to adjust burner control parameters in real time, leading to low boiler combustion efficiency.

[0003] The information disclosed in the background section of this application is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] This invention provides a boiler fuel optimization control method and system based on flame center height monitoring, which can solve the technical problems of related technologies that make it difficult to evaluate the flame combustion state in real time from multiple dimensions such as flame morphology and temperature, and make it difficult to adjust the control parameters of the burner in real time.

[0005] According to a first aspect of the present invention, a boiler fuel optimization control method based on flame center height monitoring is provided, comprising:

[0006] At multiple moments during the monitoring period, images of the flames and infrared temperature measurements inside the furnace are captured by binocular cameras installed in multiple observation holes of the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame.

[0007] Flame combustion stability indicators were determined based on flame images and infrared thermography images obtained from multiple observation holes at the flame base.

[0008] The flame morphology indicators are determined based on the flame image obtained from the observation hole in the middle of the flame.

[0009] The flame combustion status indicators are determined based on the flame image and infrared thermography image obtained from the observation hole at the tail of the flame.

[0010] Based on the flame combustion stability index, flame shape index, and flame combustion state index, determine whether the burner's control parameters need adjustment;

[0011] If the control parameters of the burners need to be adjusted, the adjustment parameters for each burner are determined based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index.

[0012] Based on the adjustment parameters, determine the adjusted fuel flow rate and adjusted air flow rate for each burner;

[0013] At the start of the next monitoring cycle, each burner is set according to the adjusted fuel flow rate and adjusted air flow rate.

[0014] According to the present invention, determining the flame combustion stability index includes:

[0015] The flame image from the observation hole at the base of the flame is segmented to obtain the first boundary between the flame area and the background area;

[0016] In the infrared thermography image from the observation hole at the base of the flame, the first isotherms of multiple temperatures were determined.

[0017] Based on the first dividing line and the first isotherm, determine the flame combustion stability index.

[0018] According to the present invention, determining the flame combustion stability index includes:

[0019] Divide the first dividing line of the flame image at each moment through the observation hole at the root of each flame into multiple first dividing points.

[0020] The multiple isotherms in the infrared thermography images of each flame root observation hole at each moment are divided into multiple second division points.

[0021] According to the formula Obtain the flame pattern contrast stability index of the flame observed through the observation hole at the root of the i-th flame. ,in, Let J be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the i-th flame. Let be the coordinates of the j-th first division point on the flame image at time t of the (i+1)-th flame root observation hole. Let be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the (i-1)-th flame. Let be the length of each equal segment of the first dividing line on the flame image at time t of the observation hole at the root of the i-th flame. To monitor the number of moments within the monitoring period, Let be the number of points in the first division, and max be the function to find the maximum value.

[0022] According to the formula The isotherm comparison stability index of the a-th temperature observed from the i-th flame root observation hole ,in, Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i+1)-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i-1)-th flame root observation hole at time t. Let be the length of each equal segment of the isotherm of the 'a' temperature in the infrared thermography image of the 'i+1' flame root observation hole at the 't' time. The number of second division points on each isotherm;

[0023] According to the formula Obtain the flame pattern time-series stability index of the flame observed through the observation hole at the root of the i-th flame. ,in, Let J be the coordinates of the j-th first division point on the flame image at the (t+1)-th moment of the observation hole at the root of the i-th flame;

[0024] According to the formula Obtain the time-series stability index of the isotherm of the a-th temperature observed from the i-th flame root observation hole. ,in, Let b be the coordinates of the second division point on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t+1.

[0025] Flame combustion stability indices are obtained based on the flame pattern stability index and flame pattern temporal stability index observed through the observation holes at the flame root of each flame, as well as the isotherm stability index and isotherm temporal stability index observed through the observation holes at the flame root of each flame.

[0026] According to the present invention, determining flame morphology indicators includes:

[0027] The flame image from the observation hole in the middle of the flame is segmented to obtain the second boundary between the flame area and the background area;

[0028] Determine the first centroid of the flame region based on the second dividing line;

[0029] Based on the calibration parameters of the RGB camera, determine the first distance between the first centroid and the location of the burner in the flame image;

[0030] Multiple third division points are evenly set along the second dividing line;

[0031] The flame shape index is determined based on the first distance and the third division point.

[0032] According to the present invention, determining flame morphology indicators includes:

[0033] According to the formula Determine the flame morphology indicators observed through the observation hole in the middle of the x-th flame. ,in, Let be the coordinates of the (y-1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the y-th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the (y+1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. For vectors with vector The angle between them The first distance corresponds to the flame image at time t through the observation hole in the middle of the x-th flame. Let x be the first theoretical distance for observation through the observation hole in the middle of the flame. Let be the number of points that form the third division, and max be the function that takes the maximum value. This refers to the number of moments within the monitoring period.

[0034] According to the present invention, determining flame combustion state indicators includes:

[0035] The flame image from the observation hole at the tail of the flame is segmented to obtain the third dividing line between the flame area and the background area;

[0036] Based on the third dividing line, determine the area ratio of the flame region to all areas captured in the flame image;

[0037] The highest temperature measured in the infrared thermography image, as well as the second isotherm at multiple temperatures, are obtained.

[0038] The flame combustion state index is determined based on the area ratio, the highest temperature, and the second isotherm.

[0039] According to the present invention, determining the flame combustion state index includes: according to the formula Determine the flame combustion state index B, where, Let be the area ratio corresponding to the flame image at time t through the observation hole at the tail of the flame. The highest temperature in the infrared thermography image at time t of the observation hole at the tail of the flame is denoted as . This represents the average of the maximum values ​​in the infrared thermography images taken at time t from multiple observation holes in the center of the flame. Let W be the maximum distance between the second isotherm of the k-th temperature and the second isotherm of the (k+1)-th temperature in the infrared thermography image taken at time t through the observation hole at the flame tail, where W is the width of the infrared thermography image and max is the function for maximizing the value. The number of second isotherms. This refers to the number of moments within the monitoring period.

[0040] According to the present invention, the training steps of the control parameter adjustment model include:

[0041] Based on the historical flame morphology index, historical flame combustion stability index, and historical flame combustion state index of the h-1th historical monitoring period, the historical flame state vector is obtained;

[0042] By adjusting the first feature extraction level of the model through control parameters, the historical flame state vector is processed to obtain historical flame state feature information.

[0043] Based on the historical fuel flow rate and historical air flow rate of each burner in the h-th historical monitoring cycle, the historical control parameter vector is obtained;

[0044] By adjusting the second feature extraction layer of the model through control parameter adjustment, the historical control parameter vector is processed to obtain historical control feature information;

[0045] Based on the first control state extraction level of the control parameter adjustment model, the historical control feature information of the h-th historical monitoring period is processed to obtain the first control state transition matrix of the h-th historical monitoring period.

[0046] Based on the second control state extraction level of the control parameter adjustment model, the historical control feature information of the h-th historical monitoring period is processed to obtain the second control state transition matrix of the h-th historical monitoring period.

[0047] The historical flame state transition vector is obtained by multiplying the first control state transition matrix with the historical flame state feature information of the (h-1)th historical monitoring cycle.

[0048] The historical control state transition vector is obtained by multiplying the second control state transition matrix with the historical control feature information of the h-th historical monitoring period.

[0049] The historical flame state transition vector and the historical control state transition vector are concatenated and processed through the first multi-layer sensing network layer to obtain the predicted flame state vector for the h-th historical monitoring cycle.

[0050] By adjusting the second multilayer sensing network layer of the control parameter adjustment model, the predicted flame state vector of the h-th historical monitoring cycle is processed to obtain the prediction adjustment parameters for the h+1-th historical monitoring cycle.

[0051] Based on the predicted adjustment parameters of the (h+1)th historical monitoring period and the historical control parameter vector of the (h)th historical monitoring period, the predicted control parameter vector of the (h+1)th historical monitoring period is obtained.

[0052] Based on the predicted control parameter vector of the (h+1)th historical monitoring cycle and the predicted flame state vector of the h-th historical monitoring cycle, as well as the control parameter adjustment model, the predicted flame state vector of the (h+1)th historical monitoring cycle is obtained.

[0053] Based on the predicted flame state vector of the (h+1)th historical monitoring cycle, the predicted flame state vector of the hth historical monitoring cycle, and the historical flame state vector of the hth historical monitoring cycle, the loss function of the control parameter adjustment model is determined.

[0054] Based on the loss function of the control parameter adjustment model, the parameter adjustment model is trained to obtain the trained control parameter adjustment model.

[0055] According to the present invention, determining the loss function of the control parameter adjustment model includes:

[0056] According to the formula Determine the loss function (LOSS) for the control parameter adjustment model, where, This is the historical flame state vector for the h-th historical monitoring period. Let h be the predicted flame state vector for the h-th historical monitoring period. This is the predicted flame state vector for the (h+1)th historical monitoring period. For preset hyperparameters less than 1, and The preset weights are defined by sim, which is the similarity calculation function.

[0057] According to a second aspect of the present invention, a boiler fuel optimization control system based on flame center height monitoring is provided, comprising:

[0058] The imaging module captures images of the flames and infrared temperature measurements inside the furnace at multiple moments during the monitoring period using binocular cameras installed in multiple observation holes in the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame.

[0059] The flame combustion stability index module determines the flame combustion stability index based on flame images and infrared thermography images from multiple flame root observation holes.

[0060] The flame morphology index module determines the flame morphology index based on the flame image viewed through the observation hole in the middle of the flame.

[0061] The flame combustion status indicator module determines the flame combustion status indicators based on the flame image from the observation hole at the tail of the flame and the infrared temperature measurement image.

[0062] The judgment module determines whether the burner's control parameters need to be adjusted based on flame combustion stability index, flame shape index, and flame combustion state index.

[0063] The adjustment parameter module determines the adjustment parameters for each burner based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index if the control parameters of the burner need to be adjusted.

[0064] The flow determination module determines the adjusted fuel flow rate and adjusted air flow rate for each burner based on the adjustment parameters.

[0065] The setting module configures each burner at the start of the next monitoring cycle based on the adjusted fuel flow and adjusted air flow.

[0066] By adopting the above technical solution, the present invention can achieve the following technical effects:

[0067] According to the present invention, based on captured flame images and infrared thermography images within the furnace, flame combustion stability indicators, flame morphology indicators, and flame combustion state indicators can be determined. This allows for determination of whether burner control parameters need adjustment, determination of adjustment parameters for each burner, and further determination of the adjusted fuel flow rate and adjusted air flow rate for each burner. Each burner is then set at the start of the next monitoring cycle. Real-time evaluation of flame combustion state from multiple dimensions such as flame morphology and temperature allows for real-time adjustment of burner control parameters, improving the accuracy and real-time performance of fuel optimization and enhancing boiler fuel combustion efficiency. Furthermore, flame images and infrared thermography images within the furnace can be captured at multiple moments within the monitoring cycle using binocular cameras installed in multiple observation ports on the boiler, providing basic data for determining whether burner control parameters need adjustment. When determining flame combustion stability indicators, multiple flame pattern comparison stability indicators, isotherm comparison stability indicators, flame pattern time-series stability indicators, and isotherm time-series stability indicators can be determined based on the first boundary line and the first isotherm, thereby determining the flame combustion stability indicators. Considering the stable combustion at the flame root, the distribution and temperature distribution of the flame at the flame root in the boiler are almost fixed in the flame image. Based on the positional differences between adjacent positions at the same time and the positional differences of the same position at different times, flame combustion stability indices are determined, improving the accuracy, objectivity, and comprehensiveness of the determination. When determining flame morphology indices, they can be based on the first distance and the third division point. Considering that under normal flame morphology, the outline of the flame observed through the observation holes in the middle of each flame should be relatively smooth, and that the centroid of the flame observed through the same observation hole in the middle of the flame remains almost unchanged at different times, flame morphology indices for the flame observed through the observation holes in the middle of each flame are determined based on the flame outline and flame position, improving the accuracy, objectivity, and comprehensiveness of the determination. When determining flame combustion state indices, they can be based on the area ratio, the highest temperature, and the second isotherm. Considering that under conditions of complete fuel combustion, the flame tail should have a smaller flame size and the combustion intensity at the flame tail should be lower than that in the flame center, values ​​describing the degree of fuel combustion completeness from three perspectives—temperature, size, and combustion intensity—are obtained, thus deriving flame combustion state indices. This improves the accuracy, objectivity, and comprehensiveness of determining flame combustion state indices. When training the control parameter adjustment model, loss functions determined based on consistency and validity can be obtained, leading to the loss function of the control parameter adjustment model. This allows for the training of the trained control parameter adjustment model, improving the accuracy and relevance of the training and enhancing the performance of the control parameter adjustment model.Considering the need to use both the predicted flame state vector from the next historical monitoring cycle and the predicted flame state vector from the current historical monitoring cycle during training, a loss function based on consistency is set according to the predicted flame state vector from the current historical monitoring cycle and the historical flame state vectors. During training, the consistency between the predicted flame state vector and the actual flame state vector is improved, thereby enhancing the accuracy of the predicted flame state vector for the next historical monitoring cycle determined by the control parameter adjustment model. Furthermore, based on the accurate predicted flame state vector for the next historical monitoring cycle and the target flame state vector, a loss function based on effectiveness is set. During training, the consistency between the predicted flame state vector for the next historical monitoring cycle and the target flame state vector is improved. This ensures that the adjustment parameters output by the control parameter adjustment model, after adjusting the burner's control parameters, can effectively reduce the degree of abnormality in flame combustion state, improving the accuracy, objectivity, and comprehensiveness of the loss function determination of the control parameter adjustment model. Further, based on the adjustment parameters, the adjusted fuel flow rate and adjusted air flow rate of each burner can be determined and set at the beginning of the next monitoring cycle. This improves the accuracy and real-time performance of fuel optimization and enhances boiler fuel combustion efficiency.

[0068] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Other features and aspects of the invention will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0069] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0070] Figure 1 An exemplary flowchart of a boiler fuel optimization control method based on flame center height monitoring according to an embodiment of the present invention is shown.

[0071] Figure 2 A flowchart for determining flame combustion stability indicators according to an embodiment of the present invention is shown as an example;

[0072] Figure 3 A block diagram of a boiler fuel optimization control system based on flame center height monitoring according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0073] 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.

[0074] 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.

[0075] Figure 1 An exemplary flowchart illustrates a boiler fuel optimization control method based on flame center height monitoring according to an embodiment of the present invention, the method comprising:

[0076] Step S1: At multiple moments during the monitoring period, images of the flames and infrared temperature measurements inside the furnace are captured by binocular cameras installed in multiple observation holes of the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame.

[0077] Step S2: Determine the flame combustion stability index based on the flame images and infrared thermography images from multiple flame root observation holes.

[0078] Step S3: Determine the flame morphology index based on the flame image from the observation hole in the middle of the flame.

[0079] Step S4: Determine the flame combustion status indicators based on the flame image from the observation hole at the tail of the flame and the infrared thermography image.

[0080] Step S5: Determine whether the control parameters of the burner need to be adjusted based on the flame combustion stability index, flame shape index, and flame combustion state index.

[0081] Step S6: If the control parameters of the burners need to be adjusted, the adjustment parameters of each burner are determined based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index.

[0082] Step S7: Determine the adjusted fuel flow rate and adjusted air flow rate for each burner based on the adjustment parameters;

[0083] Step S8: At the start of the next monitoring cycle, each burner is set according to the adjusted fuel flow rate and adjusted air flow rate.

[0084] According to embodiments of the present invention, a boiler fuel optimization control method and system based on flame center height monitoring can determine flame combustion stability indicators, flame morphology indicators, and flame combustion state indicators based on captured flame images and infrared thermography images within the furnace. This allows for determination of whether burner control parameters need adjustment, determination of adjustment parameters for each burner, and further determination of the adjusted fuel flow rate and adjusted air flow rate for each burner. At the start of the next monitoring cycle, each burner is configured. This allows for real-time evaluation of flame combustion state from multiple dimensions such as flame morphology and temperature, and real-time adjustment of burner control parameters, improving the accuracy and real-time performance of fuel optimization and enhancing boiler fuel combustion efficiency.

[0085] According to an embodiment of the present invention, in step S1, at multiple moments within the monitoring period, images of the flame and infrared temperature measurement inside the furnace are captured by binocular cameras installed in multiple observation holes on the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the flame root, flame middle, and flame tail. The monitoring period can be 30 minutes, 15 minutes, etc., and the time interval between two adjacent moments can be 30 seconds, 20 seconds, etc. When setting the observation holes, they can be based on the position of each burner. For example, an observation hole can be set at the same height as the burner, and this observation hole is located on one side of the burner, allowing observation of the flame root emitted by the burner. This observation hole is the flame root observation hole, and images of the flame root emitted by the burner and infrared temperature measurement can be captured from a horizontal perspective using the binocular cameras. A viewing port is installed at the upper part of the furnace above each burner; this is the flame center viewing port. A binocular camera can capture images of the flame center and infrared thermography from a downward-facing angle within the furnace. A viewing port is also installed at the boiler exhaust port; this is the flame tail viewing port. A binocular camera can capture images of the flame tail and infrared thermography from the flame tail within the furnace. All flame root viewing ports and flame center viewing ports are located on the same horizontal cross-section of the boiler furnace. Furthermore, flame images and infrared thermography can be captured at multiple times within the monitoring period using binocular cameras installed at multiple viewing ports on the boiler. Because the binocular camera can use a beam splitter to give the RGB camera and the infrared imager almost identical optical field of view and optical path, the pixels in the flame image and the infrared temperature measurement image corresponding to each observation hole are in one-to-one correspondence. In other words, the temperature of the corresponding pixel in the flame image can be determined based on the temperature of each pixel in the infrared temperature measurement image, which is used to determine the flame combustion stability index, flame morphology index, and flame combustion state index. This invention does not limit the duration of the monitoring period or the time interval between adjacent moments.

[0086] In this way, images of the flames and infrared temperatures inside the furnace can be captured by binocular cameras installed in multiple observation holes on the boiler at multiple times during the monitoring cycle, providing basic data for determining whether the control parameters of the burner need to be adjusted.

[0087] Figure 2 A flowchart for determining flame combustion stability indicators according to an embodiment of the present invention is shown as an example.

[0088] According to an embodiment of the present invention, in step S2, the flame combustion stability index is determined based on the flame images and infrared thermographic images from multiple flame root observation holes, including: step S21, segmenting the flame images from the flame root observation holes to obtain a first boundary line between the flame region and the background region; step S22, determining a first isotherm of multiple temperatures in the infrared thermographic images from the flame root observation holes; and step S23, determining the flame combustion stability index based on the first boundary line and the first isotherm.

[0089] According to an embodiment of the present invention, in step S21, the flame image of the observation hole at the flame root is segmented to obtain a first boundary line between the flame region and the background region. This first boundary line can be obtained by segmenting the flame image of the observation hole at the flame root using a deep learning-based semantic segmentation model (e.g., U-Net, SegNet, etc.). Based on the same processing method, multiple first boundary lines can be obtained in the flame images of each observation hole at various times.

[0090] According to an embodiment of the present invention, in step S22, a first isotherm of multiple temperatures is determined in the infrared thermography image of the observation hole at the flame root. Based on the calibration curve of the infrared thermal imager that captured the infrared thermography image, the original digital value (e.g., grayscale, radiation intensity, etc.) of each pixel in the image can be converted into an absolute temperature value (i.e., degrees Celsius), thus obtaining the temperature of each pixel. Further, methods such as thresholding or contour detection can be used to determine the isotherms of each temperature in the infrared thermography image, which are the first isotherms, where the temperature of each pixel on the first isotherm is the same. Based on the same processing method, the first isotherms of multiple temperatures in the infrared thermography images of each observation hole at each flame root at each moment can be obtained.

[0091] According to an embodiment of the present invention, in step S23, determining the flame combustion stability index based on the first boundary line and the first isotherm includes: dividing the first boundary line of the flame image at each moment of each flame root observation hole into equal parts to obtain multiple first division points; dividing multiple isotherms in the infrared thermography images at each moment of each flame root observation hole into equal parts to obtain multiple second division points; and obtaining the flame pattern comparison stability index of the flame observed by the i-th flame root observation hole according to formula (1). , (1)

[0092] in, Let J be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the i-th flame. Let be the coordinates of the j-th first division point on the flame image at time t of the (i+1)-th flame root observation hole. Let be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the (i-1)-th flame. Let be the length of each equal segment of the first dividing line on the flame image at time t of the observation hole at the root of the i-th flame. To monitor the number of moments within the monitoring period, Let be the number of points in the first division, and max be the function to find the maximum value.

[0093] According to formula (2), the isotherm comparison stability index of the a-th temperature observed by the observation hole at the i-th flame root is obtained. , (2)

[0094] in, Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i+1)-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i-1)-th flame root observation hole at time t. Let be the length of each equal segment of the isotherm of the 'a' temperature in the infrared thermography image of the 'i+1' flame root observation hole at the 't' time. The number of second division points on each isotherm;

[0095] According to formula (3), the flame pattern time-series stability index of the flame observed by the observation hole at the root of the i-th flame is obtained. , (3)

[0096] in, Let J be the coordinates of the j-th first division point on the flame image at time t+1 of the observation hole at the root of the i-th flame;

[0097] According to formula (4), the isotherm time-series stability index of the a-th temperature observed by the i-th flame root observation hole is obtained. , (4)

[0098] in, Let be the coordinates of the second division point b on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t+1; obtain the flame combustion stability index based on the flame pattern comparison stability index and flame pattern time series stability index observed by each flame root observation hole, as well as the isotherm comparison stability index and isotherm time series stability index observed by each flame root observation hole.

[0099] According to an embodiment of the present invention, the first dividing line of the flame image at each moment of each flame root observation hole is divided equally to obtain multiple first division points. For example, dividing the first dividing line of the flame image at each moment of each flame root observation hole into 19 equal parts will yield 20 first division points (including the endpoints at both ends).

[0100] According to an embodiment of the present invention, multiple isotherms in the infrared thermography images of each flame root observation hole at each moment are divided equally to obtain multiple second division points. For example, by dividing each first isotherm in the infrared thermography images of each flame root observation hole at each moment into 19 equal parts, 20 second division points (including the endpoints at both ends) can be obtained.

[0101] According to an embodiment of the present invention, in formula (1), The magnitude of the vector representing the coordinates of the j-th first division point on the flame image at time t from the observation hole at the i-th flame root, and the coordinates of the j-th first division point on the flame image at time t from the (i+1)-th flame root observation hole, can be considered as the positional difference between the two first division points. Similarly, This can represent the positional difference of the j-th first division point on the flame image at time t between the i-th and (i-1)-th flame root observation holes. Therefore, This represents the absolute value of the total positional difference between the j-th first division point on the flame image at time t of the i-th flame root observation hole and the same first division point of adjacent flame root observation holes (i.e., the previous observation hole and the next observation hole). Further, The relative value of the total position difference of the same first division point is used to eliminate the influence of the different positions of the observation holes at the root of each flame. This relative value can be used as the total position difference.

[0102] According to an embodiment of the present invention, since the flame in the boiler is generated by tangentially injecting fuel and air into a virtual circular area at the center of the furnace from multiple (e.g., four) burners, the corresponding burners have the same injection angle in the flame image from the perspective of each flame root observation hole. Therefore, when the flame is stable, the shape of the flame at the flame root in the boiler is almost fixed in the flame image. That is, the position of the same first division point observed by different flame root observation holes is almost the same. In other words, the total position difference between the first division point on the flame image at the same time of each flame root observation hole and the same first division point of adjacent flame root observation holes should be small. Therefore, the above-mentioned total position difference can be considered as the value describing the flame shape stability of the position of the j-th first division point on the flame image at the t-th time of the i-th flame root observation hole. Further, the maximum value of the value describing the flame shape stability of the positions of each first division point on the flame image at each time of the i-th flame root observation hole can be obtained. It can be used as a stability index for flame pattern comparison. The larger this index, the greater the maximum total positional difference of the same equidistant point observed by the flame root observation holes adjacent to the i-th flame root observation hole, and the worse the stability of the flame pattern observed by the i-th flame root observation hole. Based on the same processing method, the flame pattern comparison stability index of the flame observed by each flame root observation hole can be obtained. The number of flame pattern comparison stability indices is equal to the number of flame root observation holes. The number of flame root observation holes is 4, and if i=4, then i+1 is 1, i-1 is 3, that is, it can represent the two flame root observation holes adjacent to the 4th flame root observation hole.

[0103] According to an embodiment of the present invention, in formula (2), similar to formula (1), The value represents the total position difference between the b-th second division point on the isotherm of the a-th temperature in the infrared thermogram of the i-th flame root observation hole at time t after normalization and the same second division point of the adjacent flame root observation holes (i.e., the previous observation hole and the next observation hole). Since the temperature distribution of the flame at the flame root in a boiler is almost fixed under stable conditions, meaning that the positions of the same second division point of the same temperature observed by different flame root observation holes are almost identical, in other words, the total positional difference between the second division points on the isotherms of each temperature in the infrared thermography images of each flame root observation hole at each time and the same second division points of adjacent flame root observation holes (i.e., the previous observation hole and the next observation hole) should be small. Therefore, the above total positional difference can be considered as the value describing the flame temperature stability at the position of the bth second division point on the isotherm of the ath temperature in the infrared thermography image of the i-th flame root observation hole at time t. Furthermore, the maximum value of the value describing the flame temperature stability at the position of the second division point on the isotherm of the ath temperature in the infrared thermography image of the i-th flame root observation hole at each time can be obtained. It can be used as a stability index for the isotherms. The larger this index, the greater the total positional difference of the same second division point of the a-th temperature observed by the flame root observation holes adjacent to the i-th flame root observation hole, and the worse the stability of the most unstable a-th flame temperature observed by the i-th flame root observation hole. Based on the same processing method, the isotherm comparison stability index for each temperature observed by each flame root observation hole can be obtained. The number of isotherm comparison stability indices is the product of the number of flame root observation holes and the number of temperature types.

[0104] According to an embodiment of the present invention, in formula (3), This represents the position difference of the j-th first division point on the flame image at time t and time t+1 of the i-th flame root observation hole after normalization. Since the flame shape at the flame root in a boiler is relatively stable under stable conditions, and the flame shape remains almost unchanged over time, meaning the position of the same point on the flame image at different times of the same flame root observation hole is almost identical, or in other words, the position difference of the same first division point on the flame image at adjacent times of the same flame root observation hole is small, approaching 0. Therefore, the above position difference can be considered as the value describing the temporal stability of the flame pattern at the position of the j-th first division point on the flame image at time t of the i-th flame root observation hole. Furthermore, the maximum value of the value describing the temporal stability of the flame pattern at the positions of each first division point on the flame image at each time of the i-th flame root observation hole can be obtained. It can be used as a time-series stability index of the flame pattern. The larger this index, the greater the positional difference of the same point on the flame image at different times of the i-th flame root observation hole, and the worse the stability of the most unstable flame pattern observed by the i-th flame root observation hole based on time series determination. Using the same processing method, a flame pattern time series stability index for each flame observed by each flame root observation hole can be obtained. The number of flame pattern time series stability indices is equal to the number of flame root observation holes.

[0105] According to an embodiment of the present invention, in formula (4), similar to formula (3), This represents the position difference of the b-th second division point on the isotherm of temperature a in the infrared thermography images of the i-th flame root observation hole at time t and time t+1 after normalization. Since the temperature distribution of various flames at the flame root in a boiler is relatively uniform under stable flame conditions, and the temperature at each location remains almost constant over time, meaning that the position of the same point on the isotherm of the same temperature in the infrared thermography images of the same flame root observation hole at different times is almost identical. In other words, the position difference of the same second division point on the isotherm of the same temperature in the infrared thermography images of adjacent times of the same flame root observation hole is small, approaching 0. Therefore, the above position difference can be considered as a value describing the flame temperature stability at the position of the b-th second division point on the isotherm of temperature a in the infrared thermography image of the i-th flame root observation hole at time t. Furthermore, the maximum value of the flame temperature stability at the locations of the second division points of the isotherm of the a-th temperature in the infrared thermography image describing the i-th flame root observation hole at various times can be obtained. It can be used as a time-series stability index of the isotherm. The larger this index, the greater the positional difference of the same point on the isotherm of the 'a' temperature in the infrared thermography images of the 'i' flame root observation hole at different times, and the worse the stability of the most unstable 'a' flame temperature observed by the 'i' flame root observation hole based on time series determination. Using the same processing method, the isotherm time series stability index for each temperature observed by each flame root observation hole can be obtained. The number of isotherm time series stability indices is the product of the number of flame root observation holes and the number of temperature types.

[0106] According to an embodiment of the present invention, the flame combustion stability index includes the flame pattern comparison stability index and flame pattern time series stability index observed by the observation holes at the flame root of each flame, as well as the isotherm comparison stability index and isotherm time series stability index observed by the observation holes at the flame root of each flame.

[0107] In this way, multiple flame pattern comparison stability indices, isotherm comparison stability indices, flame pattern time-series stability indices, and isotherm time-series stability indices can be determined based on the first boundary line and the first isotherm, thereby determining the flame combustion stability indices. Considering that the distribution and temperature distribution of the flame at the flame root in the flame image are almost fixed when combustion at the flame root is stable, flame combustion stability indices are determined based on the positional differences between adjacent positions at the same time and the positional differences of the same position at different times, improving the accuracy, objectivity, and comprehensiveness of the determination of flame combustion stability indices.

[0108] According to an embodiment of the present invention, in step S3, determining the flame morphology index based on the flame image through the observation hole in the center of the flame includes: segmenting the flame image through the observation hole in the center of the flame to obtain a second boundary line between the flame region and the background region; determining a first centroid of the flame region based on the second boundary line; determining a first distance between the first centroid and the location of the burner in the flame image based on the calibration parameters of the RGB camera; uniformly setting a plurality of third division points on the second boundary line; and determining the flame morphology index based on the first distance and the third division points.

[0109] According to an embodiment of the present invention, similar to obtaining the first dividing line, the flame image from the observation hole in the center of the flame is segmented to obtain a second dividing line between the flame region and the background region. Since the observation hole in the center of the flame is located in the upper middle part of the boiler, the flame image captured by each observation hole in the center of the flame is a partial image of the top view of the flame. Because the flame in the boiler is generated by tangential injection of fuel and air from multiple (e.g., four) burners into a virtual circular area at the center of the furnace, when the flame is stable, the flame is nearly spherical, and the top view of the flame is nearly circular. The flame region in the flame image captured from the perspective of the observation hole in the center of the flame (i.e., the flame region in the partial area of ​​the top view of the flame) is a nearly fan-shaped region (capturing part of the flame and one burner in the flame image). The centroid of this fan-shaped region (i.e., the centroid of the flame region) is the first centroid. Further, the distance between the first centroid and the burner in the flame image in the image coordinate system, i.e., the pixel distance between them in the image, is determined as the first distance. Based on the same processing method, the first distance corresponding to the flame image at each moment of each observation hole in the middle of the flame can be determined.

[0110] According to an embodiment of the present invention, determining the flame morphology index based on the first distance and the third equidistant point includes: determining the flame morphology index of the flame observed by the observation hole at the xth flame center according to formula (5). , (5)

[0111] in, Let be the coordinates of the (y-1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the y-th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the (y+1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. For vectors with vector The angle between them The first distance corresponds to the flame image at time t through the observation hole in the middle of the x-th flame. Let x be the first theoretical distance for observation through the observation hole in the middle of the flame. Let be the number of points that form the third division, and max be the function that takes the maximum value. This refers to the number of moments within the monitoring period.

[0112] According to an embodiment of the present invention, in formula (5), For vectors with vector The angle between them can be considered as the angle between the flame shape at the position of the y-th third division point on the flame image at the t-th moment of the observation hole in the middle of the x-th flame. Furthermore, The above included angle and The ratio of the flame's shape to its angle. Since a larger angle results in a smoother flame outline, therefore... It can be considered as the value describing the smoothness of the flame outline at the location of the y-th third division point on the flame image at the t-th time through the observation hole in the middle of the x-th flame. Therefore, This can be considered as the value describing the degree of tortuosity of the flame profile at the location of the y-th third division point on the flame image at time t, which describes the position of the flame at the observation hole in the middle of the x-th flame. Furthermore, the average value of the maximum degree of tortuosity of the flame profile at each time point describing the flame image at the observation hole in the middle of the x-th flame can be obtained. This value describes the average maximum curvature of the flame outline in the flame image at each moment through the observation hole at the x-th flame center. Since the shape of a flame in a boiler is theoretically close to a regular sphere under normal flame conditions, the flame outline observed through each observation hole should be relatively smooth. Therefore, the average maximum curvature of the flame outline in the flame image at each moment through the observation hole at the x-th flame center can be considered a value describing the degree of abnormality in the flame shape observed through the observation hole at the x-th flame center, determined based on the flame outline. The larger this value is, and the closer it is to 1, the greater the curvature at the most curvature position of the flame outline in the flame image at each moment through the observation hole at the x-th flame center, and the more abnormal the flame shape observed through the observation hole at the x-th flame center.

[0113] According to an embodiment of the present invention, in formula (5), Let be the ratio of the first distance corresponding to the flame image at time t of the x-th flame center observation hole to the first theoretical distance observed by the same flame center observation hole (i.e., under the condition of uniform flame combustion, the theoretical distance between the first centroid of the flame image at each time of each flame center observation hole and the location of the burner in the same flame image). Since the centroid of the flame observed by the same flame center observation hole at different times remains almost unchanged under normal flame morphology, the position of the first centroid corresponding to the flame image at each time of the same flame center observation hole remains almost unchanged. That is, the first distance corresponding to the flame image at each time of each flame center observation hole tends to be the same as the first theoretical distance observed by the same flame center observation hole. In other words... It should approach 1. Therefore, the difference between 1 and the above ratio is... The deviation of the first distance corresponding to the flame image at time t of the observation hole at the center of the x-th flame from the first theoretical distance can be considered as the deviation of the first distance in the flame image at time t of the observation hole at the center of the x-th flame. Furthermore, the average value of the deviation of the first distance in the flame images at various times of the observation hole at the center of the x-th flame can be obtained. , can describe the average deviation of the flame height in the flame image at each moment of the x-th flame center observation hole. It can be considered as a value that describes the degree of abnormality of the flame shape observed by the x-th flame center observation hole based on the flame height. The larger the value, the closer it is to 1, the greater the average deviation of the first distance corresponding to the flame image at each moment of the x-th flame center observation hole from the first theoretical distance observed by the same flame center observation hole, and the more abnormal the flame shape observed by the x-th flame center observation hole.

[0114] According to an embodiment of the present invention, the product of the value describing the degree of flame morphological anomaly observed by the observation hole at the center of the xth flame, based on flame contour determination, and the value describing the degree of flame morphological anomaly observed by the observation hole at the center of the xth flame, based on a first distance, can be determined. It can be used as a flame morphology indicator for observing the flame through the observation hole in the middle of the xth flame. The larger this index, the more abnormal the flame outline observed by the x-th flame center observation hole, and the more abnormal the flame position, the more abnormal the flame shape observed by the x-th flame center observation hole. Based on the same processing method, the flame shape index of the flame observed by each flame center observation hole can be determined, which can describe the degree of abnormality of the flame shape from two dimensions: flame outline and height. The number of flame shape indices is equal to the number of flame center observation holes.

[0115] In this way, flame morphology indicators can be determined based on the first distance and the third division point. Considering that under normal flame morphology conditions, the flame outline observed through the observation holes in the center of each flame should be relatively smooth, and that the centroid of the flame observed through the same observation hole in the center of the flame remains almost unchanged at different times, flame morphology indicators observed through the observation holes in the center of each flame are determined based on the flame outline and flame position, thus improving the accuracy, objectivity, and comprehensiveness of flame morphology indicator determination.

[0116] According to an embodiment of the present invention, in step S4, the flame combustion state index is determined based on the flame image from the flame observation hole at the flame tail and the infrared thermography image, including: segmenting the flame image from the flame observation hole at the flame tail to obtain a third boundary line between the flame area and the background area; determining the area ratio of the flame area to all areas captured by the flame image based on the third boundary line; obtaining the highest temperature measured in the infrared thermography image and second isotherms of multiple temperatures; and determining the flame combustion state index based on the area ratio, the highest temperature, and the second isotherms.

[0117] According to an embodiment of the present invention, similar to obtaining the first dividing line, the flame image from the observation hole at the tail of the flame is segmented to obtain a third dividing line between the flame region and the background region. The ratio of the number of pixels in the flame region to the total number of pixels in the flame image is the area ratio of the flame region to all regions captured in the flame image. Similar to obtaining the first isotherm, the highest temperature measured in the infrared thermography image (i.e., the highest temperature of the flame in the image), as well as second isotherms for multiple temperatures, can be obtained.

[0118] According to an embodiment of the present invention, determining the flame combustion state index based on the area ratio, the highest temperature, and the second isotherm includes: determining the flame combustion state index B according to formula (6). (6)

[0119] in, Let be the area ratio corresponding to the flame image at time t through the observation hole at the tail of the flame. The highest temperature in the infrared thermography image at time t of the observation hole at the tail of the flame is denoted as . This represents the average of the maximum values ​​in the infrared thermography images taken at time t from multiple observation holes in the center of the flame. Let W be the maximum distance between the second isotherm of the k-th temperature and the second isotherm of the (k+1)-th temperature in the infrared thermography image taken at time t through the observation hole at the flame tail, where W is the width of the infrared thermography image and max is the function for maximizing the value. The number of second isotherms. This refers to the number of moments within the monitoring period.

[0120] According to an embodiment of the present invention, in formula (6), This represents the maximum value of the maximum distance between the second isotherms of two adjacent temperatures in the infrared thermography image of the flame tail observation hole at time t. Furthermore, the maximum value of the maximum distance between the second isotherms of two adjacent temperatures in the normalized infrared thermography image of the flame tail observation hole at time t can be obtained. Since the greater the distance between isotherms, the slower the temperature change, the higher the possibility of unburned fuel and the lower the possibility of complete combustion, the above maximum value can be considered as the value describing the completeness of combustion at time t in the combustion of the flame tail.

[0121] According to an embodiment of the present invention, in formula (6), the area ratio corresponding to the flame image at the t-th moment of the flame tail observation hole is... The value describing the size of the flame tail at time t can be considered as a value describing the degree of fuel combustion at time t. Since fuel is almost completely burned in the middle of the boiler when combustion is complete, there is almost no combustion at the tail, which is mainly used for flue gas. Therefore, only some unburned fuel is burning at the tail of the flame, meaning there is only a small flame. In other words, the area corresponding to the flame image at time t through the observation hole at the tail of the flame should be small. Therefore, the value describing the size of the flame tail at time t can be considered as a value describing the degree of fuel combustion at time t from the perspective of flame size. The larger the value, the more fuel is burned at the tail of the flame at time t, the larger the flame at the tail, and the less complete the combustion in the middle of the flame, i.e., the less complete the fuel combustion.

[0122] According to an embodiment of the present invention, in formula (6), This represents the ratio of the highest temperature in the infrared thermography image at time t from the observation port at the flame tail to the maximum value in the infrared thermography images at time t from multiple observation ports in the flame center. Since higher combustion temperatures indicate higher combustion intensity, this ratio can be considered a description of the combustion intensity at the flame tail relative to the flame center at time t. Because fuel is almost completely burned in the boiler center when combustion is complete, with almost no combustion at the tail (primarily used for flue gas), only some unburned fuel is burning at the flame tail. Therefore, the combustion intensity at the flame tail should be lower than that in the flame center. Thus, the value describing the combustion intensity at the flame tail relative to the flame center at time t can also be considered a description of the degree of fuel combustion at time t from the perspective of combustion intensity. A larger value indicates a higher combustion intensity at the flame tail relative to the flame center at time t, and less complete combustion in the flame center, meaning less complete fuel combustion.

[0123] According to an embodiment of the present invention, the product of the value describing the degree of fuel combustion at time t from the perspective of temperature, the value describing the degree of fuel combustion at time t from the perspective of flame size, and the value describing the degree of fuel combustion at time t from the perspective of combustion intensity is... This can be considered as a value describing the degree of fuel combustion at time t. Furthermore, the average value describing the degree of fuel combustion at each time point can be obtained. This value can be considered as a description of the overall degree of combustion during the fuel combustion process. It can be used as an indicator of the flame combustion state. The larger the value, the less complete the fuel combustion is, which may be due to a large difference between the flow rates of the fuel and air injected by the burner.

[0124] In this way, flame combustion state indices can be determined based on area ratio, maximum temperature, and second isotherm. Taking into account the fact that, under conditions of complete fuel combustion, the flame tail should have a smaller flame size and the combustion intensity at the flame tail should be lower than that in the flame center, values ​​describing the degree of fuel combustion completeness from three perspectives—temperature, size, and combustion intensity—are obtained, thus yielding flame combustion state indices. This improves the accuracy, objectivity, and comprehensiveness of determining flame combustion state indices.

[0125] According to an embodiment of the present invention, in step S5, it is determined whether the control parameters of the burner need to be adjusted based on the flame combustion stability index, flame morphology index, and flame combustion state index. Normal threshold values ​​for the flame combustion stability index, flame morphology index, and flame combustion state index can be set respectively (for example, a normal threshold value of 1.2 for the flame combustion stability index, 0.8 for the flame morphology index, and 0.6 for the flame combustion state index, etc.). When any one of the flame combustion stability index, flame morphology index, and flame combustion state index exceeds the normal threshold, it can be considered that there is an abnormality in fuel combustion, and the control parameters of the burner need to be adjusted.

[0126] According to an embodiment of the present invention, in step S6, if the control parameters of the burners need to be adjusted, the adjustment parameters for each burner are determined based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index. The flame combustion stability index, flame morphology index, and flame combustion state index are concatenated, and the concatenated vector is then processed by a fully connected layer to increase its dimensionality (e.g., to 128 dimensions) to obtain a flame state vector, which describes the state of flame combustion. Further, the flame combustion state vector is input into the second multilayer perceptron layer of the control parameter adjustment model (e.g., including a fully connected layer and an activation layer, where the activation layer is a network layer processed using the ReLU activation function) to obtain the adjustment parameters for each burner, such as +3%, -5%, etc.

[0127] According to an embodiment of the present invention, the training steps of the control parameter adjustment model include: obtaining a historical flame state vector based on historical flame morphology indicators, historical flame combustion stability indicators, and historical flame combustion state indicators of the (h-1)th historical monitoring period; processing the historical flame state vector through the first feature extraction layer of the control parameter adjustment model to obtain historical flame state feature information; obtaining a historical control parameter vector based on the historical fuel flow rate and historical air flow rate of each burner in the h-th historical monitoring period; processing the historical control parameter vector through the second feature extraction layer of the control parameter adjustment model to obtain historical control feature information; processing the historical control feature information of the h-th historical monitoring period according to the first control state extraction layer of the control parameter adjustment model to obtain a first control state transition matrix of the h-th historical monitoring period; processing the historical control feature information of the h-th historical monitoring period according to the second control state extraction layer of the control parameter adjustment model to obtain a second control state transition matrix of the h-th historical monitoring period; multiplying the first control state transition matrix with the historical flame state feature information of the (h-1)th historical monitoring period to obtain a historical flame state transition vector; and multiplying the second control state transition matrix with the historical flame state feature information of the h-1th historical monitoring period to obtain a historical flame state transition vector. The historical control feature information of the historical monitoring cycle is multiplied to obtain the historical control state transition vector; the historical flame state transition vector and the historical control state transition vector are concatenated and processed through the first multilayer sensing network layer to obtain the predicted flame state vector for the h-th historical monitoring cycle; the predicted flame state vector for the h-th historical monitoring cycle is processed through the second multilayer sensing network layer of the control parameter adjustment model to obtain the predicted adjustment parameters for the h+1-th historical monitoring cycle; based on the predicted adjustment parameters for the h+1-th historical monitoring cycle and the historical control parameter vector for the h-th historical monitoring cycle, the predicted flame state vector for the h+1-th historical monitoring cycle is obtained. The predicted control parameter vector for the monitoring cycle is obtained; based on the predicted control parameter vector for the (h+1)th historical monitoring cycle and the predicted flame state vector for the h-th historical monitoring cycle, and the control parameter adjustment model, the predicted flame state vector for the (h+1)th historical monitoring cycle is obtained; based on the predicted flame state vector for the (h+1)th historical monitoring cycle, the predicted flame state vector for the h-th historical monitoring cycle, and the historical flame state vector for the h-th historical monitoring cycle, the loss function of the control parameter adjustment model is determined; based on the loss function of the control parameter adjustment model, the parameter adjustment model is trained to obtain the trained control parameter adjustment model.

[0128] According to an embodiment of the present invention, flame morphology indicators, flame combustion stability indicators, and flame combustion state indicators from multiple historical monitoring periods can be obtained based on the same processing method. These are the historical flame morphology indicators, historical flame combustion stability indicators, and historical flame combustion state indicators, which are used to train the control parameter adjustment model. By concatenating the historical flame combustion stability indicators, historical flame morphology indicators, and historical flame combustion state indicators from the (h-1)th historical monitoring period, the flame state vector for the (h-1)th historical monitoring period is obtained. Further, the historical flame state vector is processed through the first feature extraction layer of the control parameter adjustment model (e.g., including multiple fully connected layers and activation layers, where the activation layer is a feature extraction layer processed using the ReLU activation function), resulting in a high-dimensional (e.g., 512-dimensional) vector, which is the historical flame state feature information, describing the true flame morphology, size, and temperature characteristics of the (h-1)th historical monitoring period.

[0129] According to an embodiment of the present invention, the historical fuel flow rate and historical air flow rate of each burner in the h-th historical monitoring cycle are concatenated, and the concatenated vector is then subjected to dimensionality-upgrading processing through a fully connected layer to obtain the control parameter vector for the h-th historical monitoring cycle, which is the historical control parameter vector. Further, similar to obtaining historical flame state feature information, the h-th historical monitoring historical control parameter vector can be processed through the second feature extraction layer of the control parameter adjustment model to obtain the h-th historical monitoring historical control feature information, which describes the true control parameters of the h-th historical monitoring burner.

[0130] According to an embodiment of the present invention, the historical control feature information of the h-th historical monitoring cycle is processed according to the first control state extraction level of the control parameter adjustment model (e.g., an extraction level including fully connected layers and reshaping layers) to obtain the first control state transition matrix of the h-th historical monitoring cycle, which can describe the changing trend of the flame state under the control action of the burner in the h-th historical monitoring cycle (i.e., predict how the flame state will change). Similar to obtaining the first control state transition matrix, the historical control feature information of the h-th historical monitoring cycle can be processed according to the second control state extraction level of the control parameter adjustment model to obtain the second control state transition matrix of the h-th historical monitoring cycle, which can describe the influence of the burner control parameters on the flame.

[0131] According to an embodiment of the present invention, by multiplying the first control state transition matrix with the historical flame state feature information of the (h-1)th historical monitoring cycle, the historical flame state feature information after transformation by the first control state matrix can be obtained, which is the historical flame state transition vector, and can describe the changed flame state (i.e., the predicted flame state of the h-th historical monitoring cycle). Similarly, by multiplying the second control state transition matrix with the historical control feature information of the h-th historical monitoring cycle, the historical control state transition vector can be obtained, which can describe the impact of the changed burner control parameters. Furthermore, the historical flame state transition vector and the historical control state transition vector are concatenated and processed through a first multilayer perceptual network layer (e.g., including fully connected layers and activation layers, where the activation layers are network layers processed using the ReLU activation function) to obtain the predicted flame state vector for the h-th historical monitoring period. This vector describes the predicted flame state for the h-th historical monitoring period after the model is processed by adjusting the control parameters. It is used to compare with the historical flame state vector for the h-th historical monitoring period (i.e., compare the actual flame state for the h-th historical monitoring period with the predicted flame state), and improve the consistency between the two during training to assist in training.

[0132] According to an embodiment of the present invention, similar to obtaining and determining the adjustment parameters of each burner, the predicted flame state vector of the h-th historical monitoring cycle is processed through the second multilayer sensing network layer of the control parameter adjustment model to obtain the predicted adjustment parameters of the h+1-th historical monitoring cycle. Then, by processing the historical control parameter vector of the h-th historical monitoring cycle of the predicted adjustment parameters, the predicted control parameter vector of the h+1-th historical monitoring cycle can be obtained, which can describe the predicted burner control parameters of the h+1-th historical monitoring cycle. Furthermore, similar to obtaining the predicted flame state vector for the h-th historical monitoring period using the historical flame state vector of the h-1th historical monitoring period, the historical control parameter vector of the h-th historical monitoring period, and the control parameter adjustment model, the predicted control parameter vector of the h+1th historical monitoring period and the predicted flame state vector of the h-th historical monitoring period can be input into the control parameter adjustment model to output the predicted flame state vector for the h+1th historical monitoring period. This vector describes the predicted flame state for the h+1th historical monitoring period after adjusting the predicted flame state for the h-th historical monitoring period using the predicted burner control parameters.

[0133] According to an embodiment of the present invention, the loss function of the control parameter adjustment model is determined based on the predicted flame state vector of the (h+1)th historical monitoring cycle, the predicted flame state vector of the hth historical monitoring cycle, and the historical flame state vector of the hth historical monitoring cycle, including: determining the loss function LOSS of the control parameter adjustment model according to formula (7). (7)

[0134] in, Let h be the historical flame state vector for the h-th historical monitoring period. Let h be the predicted flame state vector for the h-th historical monitoring period. This is the predicted flame state vector for the (h+1)th historical monitoring period. For preset hyperparameters less than 1, and The preset weights are defined by sim, which is the similarity calculation function.

[0135] According to an embodiment of the present invention, in formula (7), Let represent the similarity (e.g., cosine similarity, etc.) between the historical flame state vector of the h-th historical monitoring period and the predicted flame state vector of the h-th historical monitoring period. Therefore... This value can represent the deviation between the historical flame state vector of the h-th historical monitoring period and the predicted flame state vector of the h-th historical monitoring period. It can be considered as a value describing the deviation between the historical flame state vector of the h-th historical monitoring period and the predicted flame state vector, or it can be considered as a value describing the deviation between the predicted flame state and the actual flame state. It can be used as a loss function based on consistency determination. During training, this loss function is reduced to decrease the deviation between the historical flame state vector of the h-th historical monitoring period and the predicted flame state vector, that is, to reduce the deviation between the predicted flame state and the actual flame state, thereby improving the accuracy of prediction. This assists in the training of the control parameter adjustment model, which is beneficial to improving the accuracy of the control parameter adjustment model in predicting the flame state vector of the next monitoring period based on the flame state vector of the current monitoring period. Furthermore, based on the accurately predicted flame state vector of the next monitoring period, the control parameter adjustment model is trained to obtain the predicted adjustment parameters for the next monitoring period. In other words, based on the accurately predicted flame state of the next monitoring period, the burner adjustment parameters for the next monitoring period can be predicted, improving the accuracy of burner control parameter determination.

[0136] According to an embodiment of the present invention, in formula (7), The magnitude of the predicted flame state vector for the (h+1)th historical monitoring period. Since the predicted flame state vector for the (h+1)th historical monitoring period is composed of the predicted flame combustion stability index, predicted flame morphology index, and predicted flame combustion state index for the (h+1)th historical monitoring period, and can describe the flame combustion state, the magnitude of the predicted flame state vector for the (h+1)th historical monitoring period can be considered a value describing the degree of anomalousness of the predicted flame combustion state for the (h+1)th historical monitoring period. Because the larger the flame combustion stability index, flame morphology index, and flame combustion state index, the more anomalous the flame combustion state, the larger the value describing the degree of anomalousness of the predicted flame combustion state for the (h+1)th historical monitoring period, the more anomalous the predicted flame combustion state for the (h+1)th historical monitoring period. Similarly, This can be considered as a value describing the degree of abnormality in the combustion state of the flame in the h-th historical monitoring period. Further, the product of the preset hyperparameter (e.g., 0.9, 0.95, etc.) and the value describing the abnormality in the combustion state of the flame in the h-th historical monitoring period can be considered as a target value describing the predicted degree of abnormality in the combustion state of the flame in the (h+1)-th historical monitoring period. After processing with the hyperparameter, the target value describing the predicted degree of abnormality in the combustion state of the flame in the (h+1)-th historical monitoring period is lower than the value describing the abnormality in the combustion state of the flame in the h-th historical monitoring period. That is, by approaching this target value, the degree of abnormality in the combustion state of the flame can be reduced. The difference between this target value and the value describing the predicted degree of abnormality in the combustion state of the flame in the (h+1)-th historical monitoring period can be considered as the deviation between the predicted degree of abnormality in the combustion state of the flame in the (h+1)-th historical monitoring period and the target value, and can be used as a loss function determined based on effectiveness. During training, the loss function is reduced to decrease the deviation between the predicted anomaly level of the flame combustion state over the (h+1)th historical monitoring period and the target value. This allows, after training, the adjustment parameters output by the control parameter adjustment model to modify the burner's control parameters, further reducing the magnitude of the value describing the anomaly level of the flame combustion state. In other words, it effectively reduces the anomaly level of the flame combustion state and improves its combustion. Furthermore, by presetting weights... and By weighted summing the loss function determined based on consistency and the loss function determined based on effectiveness, the loss function of the control parameter adjustment model can be obtained. That is, while improving the prediction accuracy of the predicted flame state vector for the next monitoring cycle, the control parameter adjustment model further constrains the training of the second multilayer sensing network layer by reducing the anomaly level of the predicted flame state vector for the next monitoring cycle (i.e., making the predicted flame state vector for the next monitoring cycle not only accurate but also less anomaly-prone). This allows the second multilayer sensing network layer to output prediction adjustment parameters that reduce the anomaly level of the flame. The second multilayer sensing network layer can be trained through bi-cycle prediction and constraint.

[0137] According to an embodiment of the present invention, the parameter adjustment model is trained based on the loss function of the control parameter adjustment model to obtain the trained control parameter adjustment model. The control parameter adjustment model is trained by adjusting its parameters using gradient descent based on the loss function of the control parameter adjustment model. After multiple training iterations (i.e., training using predicted flame state vectors, predicted flame state vectors, and historical flame state vectors from multiple historical monitoring periods), the training is completed, and the trained control parameter adjustment model is obtained.

[0138] In this way, when training the control parameter adjustment model, loss functions determined based on consistency and loss functions determined based on effectiveness can be obtained, thereby obtaining the loss function of the control parameter adjustment model and training the trained control parameter adjustment model, which improves the accuracy and relevance of training and enhances the performance of the control parameter adjustment model. Considering the need to use the predicted flame state vectors of the next historical monitoring cycle and the current historical monitoring cycle during training, a loss function based on consistency is set according to the predicted flame state vectors of the current historical monitoring cycle and the historical flame state vectors. During training, the consistency between the predicted flame state vector and the actual flame state vector is improved to enhance the accuracy of the predicted flame state vector for the next historical monitoring cycle determined by the control parameter adjustment model. Furthermore, based on the accurate predicted flame state vector for the next historical monitoring cycle and the target flame state vector, a loss function based on effectiveness is set. During training, the consistency between the predicted flame state vector for the next historical monitoring cycle and the target flame state vector is improved. This ensures that the adjustment parameters output by the control parameter adjustment model, after adjusting the burner's control parameters, can effectively reduce the abnormality of the flame combustion state, thus improving the accuracy, objectivity, and comprehensiveness of the loss function determination of the control parameter adjustment model.

[0139] According to an embodiment of the present invention, in step S7, the adjusted fuel flow rate and adjusted air flow rate of each burner are determined according to the adjustment parameters. For example, if the adjustment parameters for a certain burner are fuel flow rate +3% and air flow rate +1%, and the original fuel flow rate is 1500 m³ / h and the original air flow rate is 17000 m³ / h, then the adjusted fuel flow rate is (1+3%)×1500 m³ / h, i.e., 1545 m³ / h, and the adjusted air flow rate is (1+1%)×17000 m³ / h, i.e., 17170 m³ / h.

[0140] According to an embodiment of the present invention, in step S8, at the beginning of the next monitoring cycle, each burner is set according to the adjusted fuel flow rate and adjusted air flow rate of each burner to optimize the flame combustion state and make the fuel burn more stably and completely.

[0141] In this way, the adjusted fuel flow and air flow of each burner can be determined based on the adjustment parameters, and the settings for each burner can be configured at the beginning of the next monitoring cycle. This improves the accuracy and real-time performance of fuel optimization and enhances boiler fuel combustion efficiency.

[0142] According to embodiments of the present invention, a boiler fuel optimization control method and system based on flame center height monitoring can determine flame combustion stability indicators, flame morphology indicators, and flame combustion state indicators based on captured flame images and infrared thermography images within the furnace. This allows for determination of whether burner control parameters need adjustment, determination of adjustment parameters for each burner, and further determination of the adjusted fuel flow rate and adjusted air flow rate for each burner. At the start of the next monitoring cycle, each burner is configured. The flame combustion state can be evaluated in real time from multiple dimensions such as flame morphology and temperature, and burner control parameters can be adjusted in real time, improving the accuracy and real-time performance of fuel optimization and enhancing boiler fuel combustion efficiency. Furthermore, at multiple moments within the monitoring cycle, flame images and infrared thermography images within the furnace can be captured by binocular cameras installed in multiple observation ports on the boiler, providing basic data for determining whether burner control parameters need adjustment. When determining flame combustion stability indicators, multiple flame pattern comparison stability indicators, isotherm comparison stability indicators, flame pattern time-series stability indicators, and isotherm time-series stability indicators can be determined based on a first boundary line and a first isotherm, thereby determining the flame combustion stability indicators. Considering the stable combustion at the flame root, the distribution and temperature distribution of the flame at the flame root in the boiler are almost fixed in the flame image. Based on the positional differences between adjacent positions at the same time and the positional differences of the same position at different times, flame combustion stability indices are determined, improving the accuracy, objectivity, and comprehensiveness of the determination. When determining flame morphology indices, they can be based on the first distance and the third division point. Considering that under normal flame morphology, the outline of the flame observed through the observation holes in the middle of each flame should be relatively smooth, and that the centroid of the flame observed through the same observation hole in the middle of the flame remains almost unchanged at different times, flame morphology indices for the flame observed through the observation holes in the middle of each flame are determined based on the flame outline and flame position, improving the accuracy, objectivity, and comprehensiveness of the determination. When determining flame combustion state indices, they can be based on the area ratio, the highest temperature, and the second isotherm. Considering that under conditions of complete fuel combustion, the flame tail should have a smaller flame size and the combustion intensity at the flame tail should be lower than that in the flame center, values ​​describing the degree of fuel combustion completeness from three perspectives—temperature, size, and combustion intensity—are obtained, thus deriving flame combustion state indices. This improves the accuracy, objectivity, and comprehensiveness of determining flame combustion state indices. When training the control parameter adjustment model, loss functions determined based on consistency and validity can be obtained, leading to the loss function of the control parameter adjustment model. This allows for the training of the trained control parameter adjustment model, improving the accuracy and relevance of the training and enhancing the performance of the control parameter adjustment model.Considering the need to use both the predicted flame state vector from the next historical monitoring cycle and the predicted flame state vector from the current historical monitoring cycle during training, a loss function based on consistency is set according to the predicted flame state vector from the current historical monitoring cycle and the historical flame state vectors. During training, the consistency between the predicted flame state vector and the actual flame state vector is improved, thereby enhancing the accuracy of the predicted flame state vector for the next historical monitoring cycle determined by the control parameter adjustment model. Furthermore, based on the accurate predicted flame state vector for the next historical monitoring cycle and the target flame state vector, a loss function based on effectiveness is set. During training, the consistency between the predicted flame state vector for the next historical monitoring cycle and the target flame state vector is improved. This ensures that the adjustment parameters output by the control parameter adjustment model, after adjusting the burner's control parameters, can effectively reduce the degree of abnormality in flame combustion state, improving the accuracy, objectivity, and comprehensiveness of the loss function determination of the control parameter adjustment model. Further, based on the adjustment parameters, the adjusted fuel flow rate and adjusted air flow rate of each burner can be determined and set at the beginning of the next monitoring cycle. This improves the accuracy and real-time performance of fuel optimization and enhances boiler fuel combustion efficiency.

[0143] Figure 3 An exemplary block diagram of a boiler fuel optimization control system based on flame center height monitoring according to an embodiment of the present invention is shown, the system comprising:

[0144] The imaging module captures images of the flames and infrared temperature measurements inside the furnace at multiple moments during the monitoring period using binocular cameras installed in multiple observation holes in the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame.

[0145] The flame combustion stability index module determines the flame combustion stability index based on flame images and infrared thermography images from multiple flame root observation holes.

[0146] The flame morphology index module determines the flame morphology index based on the flame image viewed through the observation hole in the middle of the flame.

[0147] The flame combustion status indicator module determines the flame combustion status indicators based on the flame image from the observation hole at the tail of the flame and the infrared temperature measurement image.

[0148] The judgment module determines whether the burner's control parameters need to be adjusted based on flame combustion stability index, flame shape index, and flame combustion state index.

[0149] The adjustment parameter module determines the adjustment parameters for each burner based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index if the control parameters of the burner need to be adjusted.

[0150] The flow determination module determines the adjusted fuel flow rate and adjusted air flow rate for each burner based on the adjustment parameters.

[0151] The setting module configures each burner at the start of the next monitoring cycle based on the adjusted fuel flow and adjusted air flow.

[0152] 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.

[0153] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

[0154] 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. A boiler fuel optimization control method based on flame center height monitoring, characterized in that, include: At multiple moments during the monitoring period, images of the flames and infrared temperature measurements inside the furnace are captured by binocular cameras installed in multiple observation holes of the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame. Flame combustion stability indicators were determined based on flame images and infrared thermography images obtained from multiple observation holes at the flame base. The flame morphology indicators are determined based on the flame image obtained from the observation hole in the middle of the flame. The flame combustion status indicators are determined based on the flame image and infrared thermography image obtained from the observation hole at the tail of the flame. Based on the flame combustion stability index, flame shape index, and flame combustion state index, determine whether the burner's control parameters need adjustment; If the control parameters of the burners need to be adjusted, the adjustment parameters for each burner are determined based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index. Based on the adjustment parameters, determine the adjusted fuel flow rate and adjusted air flow rate for each burner; At the start of the next monitoring cycle, each burner is set according to the adjusted fuel flow and adjusted air flow. Based on flame images and infrared thermography images obtained from multiple observation holes at the flame base, flame combustion stability indicators were determined, including: The flame image from the observation hole at the base of the flame is segmented to obtain the first boundary between the flame area and the background area; In the infrared thermography image from the observation hole at the base of the flame, the first isotherms of multiple temperatures were determined. Determine the flame combustion stability index based on the first dividing line and the first isotherm; Based on the flame image viewed through the observation hole in the middle of the flame, determine the flame morphology indicators, including: The flame image from the observation hole in the middle of the flame is segmented to obtain the second boundary between the flame area and the background area; Determine the first centroid of the flame region based on the second dividing line; Based on the calibration parameters of the RGB camera, determine the first distance between the first centroid and the location of the burner in the flame image; Multiple third division points are evenly set along the second dividing line; The flame shape index is determined based on the first distance and the third division point; Based on the flame image obtained from the observation port at the tail of the flame and the infrared thermography image, the flame combustion status indicators are determined, including: The flame image from the observation hole at the tail of the flame is segmented to obtain the third dividing line between the flame area and the background area; Based on the third dividing line, determine the area ratio of the flame region to all areas captured in the flame image; The highest temperature measured in the infrared thermography image, as well as the second isotherm at multiple temperatures, are obtained. The flame combustion state index is determined based on the area ratio, the highest temperature, and the second isotherm. The training steps of the control parameter adjustment model include: Based on the historical flame morphology index, historical flame combustion stability index, and historical flame combustion state index of the h-1th historical monitoring period, the historical flame state vector is obtained; By adjusting the first feature extraction level of the model through control parameters, the historical flame state vector is processed to obtain historical flame state feature information. Based on the historical fuel flow rate and historical air flow rate of each burner in the h-th historical monitoring cycle, the historical control parameter vector is obtained; By adjusting the second feature extraction layer of the model through control parameter adjustment, the historical control parameter vector is processed to obtain historical control feature information; Based on the first control state extraction level of the control parameter adjustment model, the historical control feature information of the h-th historical monitoring period is processed to obtain the first control state transition matrix of the h-th historical monitoring period. Based on the second control state extraction level of the control parameter adjustment model, the historical control feature information of the h-th historical monitoring period is processed to obtain the second control state transition matrix of the h-th historical monitoring period. The historical flame state transition vector is obtained by multiplying the first control state transition matrix with the historical flame state feature information of the (h-1)th historical monitoring cycle. The historical control state transition vector is obtained by multiplying the second control state transition matrix with the historical control feature information of the h-th historical monitoring period. The historical flame state transition vector and the historical control state transition vector are concatenated and processed through the first multi-layer sensing network layer to obtain the predicted flame state vector for the h-th historical monitoring cycle. By adjusting the second multilayer sensing network layer of the control parameter adjustment model, the predicted flame state vector of the h-th historical monitoring cycle is processed to obtain the prediction adjustment parameters for the h+1-th historical monitoring cycle. Based on the predicted adjustment parameters of the (h+1)th historical monitoring period and the historical control parameter vector of the (h)th historical monitoring period, the predicted control parameter vector of the (h+1)th historical monitoring period is obtained. Based on the predicted control parameter vector of the (h+1)th historical monitoring cycle and the predicted flame state vector of the h-th historical monitoring cycle, as well as the control parameter adjustment model, the predicted flame state vector of the (h+1)th historical monitoring cycle is obtained. Based on the predicted flame state vector of the (h+1)th historical monitoring cycle, the predicted flame state vector of the hth historical monitoring cycle, and the historical flame state vector of the hth historical monitoring cycle, the loss function of the control parameter adjustment model is determined. Based on the loss function of the control parameter adjustment model, the parameter adjustment model is trained to obtain the trained control parameter adjustment model.

2. The boiler fuel optimization control method based on flame center height monitoring according to claim 1, characterized in that, Based on the first dividing line and the first isotherm, flame combustion stability indicators are determined, including: Divide the first dividing line of the flame image at each moment through the observation hole at the root of each flame into multiple first dividing points. The multiple isotherms in the infrared thermography images of each flame root observation hole at each moment are divided into multiple second division points. According to the formula Obtain the flame pattern contrast stability index of the flame observed through the observation hole at the root of the i-th flame. ,in, Let J be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the i-th flame. Let be the coordinates of the j-th first division point on the flame image at time t of the (i+1)-th flame root observation hole. Let be the coordinates of the j-th first division point on the flame image at time t of the observation hole at the root of the (i-1)-th flame. Let be the length of each equal segment of the first dividing line on the flame image at time t of the observation hole at the root of the i-th flame. To monitor the number of moments within the monitoring period, Let be the number of points in the first division, and max be the function to find the maximum value. According to the formula The isotherm comparison stability index of the a-th temperature observed from the i-th flame root observation hole ,in, Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i+1)-th flame root observation hole at time t. Let be the coordinates of the b-th second division point on the isotherm of the a-th temperature in the infrared thermography image of the (i-1)-th flame root observation hole at time t. Let be the length of each equal segment of the isotherm of the 'a' temperature in the infrared thermography image of the 'i+1' flame root observation hole at the 't' time. The number of second division points on each isotherm; According to the formula Obtain the flame pattern time-series stability index of the flame observed through the observation hole at the root of the i-th flame. ,in, Let J be the coordinates of the j-th first division point on the flame image at the (t+1)-th moment of the observation hole at the root of the i-th flame; According to the formula Obtain the time-series stability index of the isotherm of the a-th temperature observed from the i-th flame root observation hole. ,in, Let b be the coordinates of the second division point on the isotherm of the a-th temperature in the infrared thermography image of the i-th flame root observation hole at time t+1. Flame combustion stability indices are obtained based on the flame pattern stability index and flame pattern temporal stability index observed through the observation holes at the flame root of each flame, as well as the isotherm stability index and isotherm temporal stability index observed through the observation holes at the flame root of each flame.

3. The boiler fuel optimization control method based on flame center height monitoring according to claim 1, characterized in that, Based on the first distance and the third division point, flame morphology indicators are determined, including: According to the formula Determine the flame morphology indicators observed through the observation hole in the middle of the x-th flame. ,in, Let be the coordinates of the (y-1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the y-th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. Let be the coordinates of the (y+1)th third division point on the flame image at time t of the observation hole in the middle of the x-th flame. For vectors with vector The angle between them The first distance corresponds to the flame image at time t through the observation hole in the middle of the x-th flame. Let x be the first theoretical distance for observation through the observation hole in the middle of the flame. Let be the number of points that divide the text into third equal parts, and max be the function that takes the maximum value. This refers to the number of moments within the monitoring period.

4. The boiler fuel optimization control method based on flame center height monitoring according to claim 1, characterized in that, Based on the area ratio, the highest temperature, and the second isotherm, flame combustion state indicators are determined, including: According to the formula Determine the flame combustion state index B, where, Let be the area ratio corresponding to the flame image at time t through the observation hole at the tail of the flame. The highest temperature in the infrared thermography image at time t of the observation hole at the tail of the flame is denoted as . This represents the average of the maximum values ​​in the infrared thermography images taken at time t from multiple observation holes in the center of the flame. Let W be the maximum distance between the second isotherm of the k-th temperature and the second isotherm of the (k+1)-th temperature in the infrared thermography image taken at time t through the observation hole at the flame tail, where W is the width of the infrared thermography image and max is the function for maximizing the value. The number of second isotherms. This refers to the number of moments within the monitoring period.

5. The boiler fuel optimization control method based on flame center height monitoring according to claim 1, characterized in that, Based on the predicted flame state vector of the (h+1)th historical monitoring period, the predicted flame state vector of the hth historical monitoring period, and the historical flame state vector of the hth historical monitoring period, the loss function of the control parameter adjustment model is determined, including: According to the formula Determine the loss function (LOSS) for the control parameter adjustment model, where, This is the historical flame state vector for the h-th historical monitoring period. Let h be the predicted flame state vector for the h-th historical monitoring period. This is the predicted flame state vector for the (h+1)th historical monitoring period. For preset hyperparameters less than 1, and The preset weights are defined by sim, which is the similarity calculation function.

6. A boiler fuel optimization control system based on flame center height monitoring, used to execute the method as described in any one of claims 1-5, characterized in that, include: The imaging module captures images of the flames and infrared temperature measurements inside the furnace at multiple moments during the monitoring period using binocular cameras installed in multiple observation holes in the boiler. The binocular cameras include an RGB camera and an infrared thermal imager. The observation holes include observation holes at the root of the flame, observation holes in the middle of the flame, and observation holes at the tail of the flame. The flame combustion stability index module determines the flame combustion stability index based on flame images and infrared thermography images from multiple flame root observation holes. The flame morphology index module determines the flame morphology index based on the flame image viewed through the observation hole in the middle of the flame. The flame combustion status indicator module determines the flame combustion status indicators based on the flame image from the observation hole at the tail of the flame and the infrared temperature measurement image. The judgment module determines whether the burner's control parameters need to be adjusted based on flame combustion stability index, flame shape index, and flame combustion state index. The adjustment parameter module determines the adjustment parameters for each burner based on the trained control parameter adjustment model, flame combustion stability index, flame morphology index, and flame combustion state index if the control parameters of the burner need to be adjusted. The flow determination module determines the adjusted fuel flow rate and adjusted air flow rate for each burner based on the adjustment parameters. The setting module configures each burner at the start of the next monitoring cycle based on the adjusted fuel flow and adjusted air flow.

Images