Combustor flame combustion stability monitoring method based on flame image features

Abstract

The invention discloses a combustor flame combustion stability monitoring method based on flame image features, and belongs to the technical field of flame state detection. Comprising the following steps: preprocessing a flame image acquired by a camera in real time; flame characteristic parameters are extracted; calculating a speed mean value and a speed direction mean value of each frame of pixel point in the flame image; measuring the dispersion degree of the data set of the flame image by using the standard deviation; introducing variation parameters, and eliminating the dimensional influence of the data; dividing confidence intervals of the data set to obtain stability indexes of the flame characteristic parameters; and according to the specific numerical value of each frame of flame image, endowing different flame characteristic parameters with corresponding weights by utilizing a CRITIC method, and calculating the weighted sum of all the flame characteristic parameters to obtain the flame combustion stability of the combustor. The method has the advantages of being high in response speed, high in monitoring precision, not limited by working conditions and the like, and provides powerful guarantee for safe and efficient operation of the industrial combustor.

Description

Technical Field

[0001] The invention relates to the technical field of flame state detection, and in particular to a burner flame combustion stability monitoring method based on flame image features. Background Art

[0002] In industrial processes, burners are widely used in equipment such as boilers, heating furnaces and industrial kilns, and ensuring their safe and efficient operation is crucial. The stability of the flame directly affects the performance of the burner. If the flame is unstable, such as flameout or flameout, it will not only lead to a decrease in combustion efficiency, but may also cause safety accidents. Therefore, real-time monitoring and evaluation of the stability of the burner flame is of great significance to improving the performance of the burner and ensuring safe operation.

[0003] At present, many flame monitoring methods mainly rely on means such as flame detectors, temperature sensors or pressure sensors. Although these methods can provide certain flame state information, they have problems such as slow response speed, susceptibility to interference in complex combustion environments, and high installation and maintenance costs of sensors. With the advancement of image processing and computer vision technology, monitoring methods based on flame image features have gradually attracted attention. Flame images not only contain rich information such as flame brightness, color and shape, but also reflect the dynamic change process of flames. Therefore, by analyzing flame images, the stability of flames can be monitored more comprehensively and accurately. However, the existing flame stability detection methods based on flame images extract fewer flame features, such as flame root area, brightness and flame contour. These methods have many limitations. For example, when the camera is not installed near the burner, some flame features such as the flame root area cannot be extracted. This method can only reflect the flame stability state from a single angle, which has a certain impact on the analysis of flame stability. In addition, there are some methods that combine flame image features with machine learning to analyze flame stability, but there are problems such as complex algorithms that cannot reflect the flame stability state in real time.

[0004] Therefore, a new burner flame combustion stability monitoring method based on flame image features is needed to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a burner flame combustion stability monitoring method based on flame image features, comprising the following steps:

[0006] Step 1: Perform preprocessing operations such as grayscale conversion, filtering and denoising, image enhancement, and flame area segmentation on the flame image collected by the camera in real time;

[0007] Step 2: Extract flame characteristic parameters;

[0008] Step 3: Calculate the mean value of the speed magnitude and the mean value of the speed direction of each pixel in the flame image;

[0009] Step 4: Use the standard deviation to measure the discreteness of the flame image data set;

[0010] Step 5: Introduce variation parameters to eliminate the dimensionality effect of data;

[0011] Step 6: Divide the confidence interval of the data set to obtain the stability index of the flame characteristic parameters;

[0012] Step 7: According to the specific values ​​of the flame image of each frame, the CRITIC method is used to assign corresponding weights to different flame characteristic parameters, and the weighted sum of all flame characteristic parameters is calculated to obtain the burner flame combustion stability.

[0013] The flame characteristic parameters in step 2 include average brightness, unevenness, oscillation frequency, flame temperature, and flame light flow vector velocity magnitude and direction.

[0014] The stability index of the flame characteristic parameters in step 6 is as follows:

[0015]

[0016] In the formula, δ par is the flame characteristic parameter stability index, σ x is the standard deviation of the feature parameter data set, μ x is the mean of the data set, ν var_out is the variation parameter of the data outside the 95% confidence interval, ν var_in is the variation parameter of the data within the 95% confidence interval, and w1 and w2 are weight coefficients.

[0017] The beneficial effects of the present invention are:

[0018] The method of the invention can realize reliable and accurate online monitoring of boiler combustion stability, which helps power plant operators to timely and reasonably adjust parameters such as coal quantity and air-to-coal ratio to improve combustion stability; and can guide power plants to explore the deep adjustment space, which has important practical significance for the economy, safety and stability of power plants. In addition, the method has the advantages of fast response speed, high monitoring accuracy, and no restrictions on working conditions, which provides a strong guarantee for the safe and efficient operation of industrial burners. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a burner flame combustion stability monitoring method based on flame image features of the present invention;

[0020] In the figure: 1-flame, 2-burner, 3-fire detection probe, 4-PLC, 5-industrial computer;

[0021] Figure 2Design diagram for burner flame stability monitoring system;

[0022] Figure 3 It is the flow chart of flame characteristic parameter stability index calculation;

[0023] Figure 4 The curves of coal feeding amount and primary air volume changing with time;

[0024] Figure 5 is the curve of unit load changing with time;

[0025] Figure 6 This is the curve of flame stability index changing with time. DETAILED DESCRIPTION

[0026] The present invention proposes a burner flame combustion stability monitoring method based on flame image features, which will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0027] Figure 1 It is a flow chart of a burner flame combustion stability monitoring method based on flame image features of the present invention; Figure 2 This is the design diagram of the burner flame stability monitoring system. The burner flame stability monitoring system consists of an image acquisition system, a control system, and an image analysis system. The PLC4 in the control system controls the flame detection probe 3 in the image acquisition system to continuously shoot the flame 1 image of the burner 2, and then the image is transmitted to the industrial computer 5 in the analysis cabinet in the data image analysis system through the optical fiber, and then the flame image is saved and the stability analysis of the real-time flame image is performed. The details are as follows:

[0028] Firstly, the flame image captured by the camera in real time was preprocessed by grayscale conversion, filtering and denoising, image enhancement and flame area segmentation, and multiple feature parameters were extracted from the processed image.

[0029] Average brightness I:

[0030]

[0031] In the formula, I is the average brightness, R f is the flame luminous area, and A is the number of pixels in the luminous area.

[0032] Unevenness NU:

[0033]

[0034] Where NU is the unevenness, N p is the total number of pixels in the luminous area, I(k) is the image grayscale level 0 to 255, n(k) is the number of pixels corresponding to the grayscale I(k), and I f is the average brightness value before normalization.

[0035] Oscillation frequency F:

[0036]

[0037] Where F is the oscillation frequency, f i is the i-th frequency component in the spectrum, p i is the power density of the ith frequency component, and N is the total number of frequency components.

[0038] Flame temperature: According to Planck's radiation law, the two-color temperature measurement formula can be derived as shown below.

[0039]

[0040] The average temperature formula is shown below.

[0041]

[0042] Where T(i,j) is the flame temperature at the flame image coordinate ((i,j), T av is the average flame temperature, C2 is the second Planck constant, λ1 is the peak wavelength of the green spectrum, λ2 is the peak wavelength of the red spectrum, G1 and G2 are the brightness values ​​corresponding to the R and G components of the image, respectively, and S is the device factor, which can be obtained through blackbody furnace calibration. Since the camera sensor has a higher spectral sensitivity to the G and R channels than the B channel, a better signal-to-noise ratio can be achieved, so the R and G images are selected here for temperature calculation.

[0043] Flame optical flow vector speed magnitude and direction: Calculate the average speed magnitude and speed direction of each pixel in the flame image. Starting from the basic formula of the HS optical flow method, combined with the pyramid iteration method, according to the brightness conservation assumption of the optical flow method, the following optical flow constraint equation is listed for each pixel of the image:

[0044] I(x+dx,y+dy,t+dt)=I(x,y,t)

[0045] Where I(x, y, t) represents the grayscale value of the image position at time t.

[0046] After Taylor expansion of the above formula, we can simplify it to get:

[0047] I x u+I y v+I t =0

[0048] Among them I x ,I y ,I tIt represents the gradient of the gray value of a pixel in the x, y and time directions, and u and v represent the speed of the pixel in the x and y axis directions respectively.

[0049] The basic idea of ​​the HS optical flow method is to solve the optical flow field (u, v) by minimizing the following energy function:

[0050]

[0051] in: is a regularization term. The larger the value, the less smooth the velocity field is. α is a smoothing parameter.

[0052] According to Lagrange equation:

[0053]

[0054] Combining the above two equations, we can get

[0055]

[0056] Where Δ is the Laplace operator,

[0057] Taking pixel (i, j) as an example, Δu can be expressed as

[0058] Δu i,j =(u i+1,j -u i,j )-(u i,j -u i-1,j )+(u i,j+1 -u i,j )-(u i,j -u i,j-1 )

[0059] =u i+1,j +u i-1,j +u i,j+1 +u i,j-1 -4u i,j

[0060] Discretize the equation and simplify it to get

[0061]

[0062] The above method is only applicable to the case where there is a small displacement between two consecutive frames, but the flame image actually has a large displacement, so the pyramid iteration algorithm is introduced to solve the case of large displacement. Starting from the top layer of the pyramid, the velocity field is calculated layer by layer downward. Here, the image deformation technology is needed to first enlarge the velocity field matrix obtained from the top layer to the same size as the middle layer of the pyramid, and then enlarge the scalar value of the velocity field according to the corresponding proportion, that is, to obtain the initial value of the velocity field iteration of the middle layer of the pyramid. Then perform image deformation operation according to Reducing the difference between I1 and I2, we get and They represent the image after deformation operation, satisfying the small displacement condition of two consecutive frames of the optical flow method.

[0063] Get the new iteration formula:

[0064]

[0065] For each pixel (i, j), the velocity vector n i =(u i ,v i )'s speed d i With the movement angle θ i They are:

[0066]

[0067]

[0068] Select the mean of the velocity magnitude and direction of all pixels in each frame of the image.

[0069] Mean speed:

[0070]

[0071] Direction size mean:

[0072]

[0073] Where n is the total number of pixels in each frame of flame image.

[0074] In data analysis, the standard deviation is usually used to measure the degree of dispersion of a data set. When the standard deviation of a set of data is low, it indicates that the data is close to the average value, the data distribution is relatively uniform, and the combustion situation tends to be stable; when the standard deviation of the data is high, it indicates that the data distribution is in a larger range of values, and the combustion is unstable. The standard deviation of the parameter is defined as follows:

[0075]

[0076] In the formula, σ is the standard deviation of the data set, X i is the i-th data point in the data set, μ is the mean of the data set, and n represents the total number of points in the data set.

[0077] In order to eliminate the dimensionality effect of the data and make the comparison between different data sets fairer, the variation parameter is introduced. The variation parameter is defined as the ratio of the standard deviation to the mean, as shown in the following formula. It is usually used to measure the degree of variation of the data, characterize the degree of dispersion of the data, and reflect the degree of variation of the group of data relative to its mean, that is, the relative degree of variation of the group of data. The larger the value, the more dispersed the group of data is in the entire range, and the data changes are not stable; conversely, it means that the distribution of the group of data is relatively concentrated, and the data changes are small.

[0078]

[0079] Where, ν var is the variation parameter of this group of data, σ is the standard deviation of this group of data, and μ is the mean of this group of data.

[0080] Figure 3 This is the calculation flow chart of the flame characteristic parameter stability index. Considering the data reliability and sensitivity, the data set is divided into two parts according to the confidence interval: one part is the data within the 95% confidence interval, and the other part is the data outside the 95% confidence interval. The standard deviation σ of the data within the 95% confidence interval is calculated respectively. in With mean μ in , get the variation parameter ν of the data within the confidence interval var_in The data outside the 95% confidence interval are divided into two parts according to the data size, and their variation parameters are calculated respectively. The average value of the two is taken as the variation parameter ν of the data outside the 95% confidence interval. var_out The flame characteristic parameter stability index is defined as follows:

[0081]

[0082] In the formula, δ par is the flame characteristic parameter stability index, σ x is the standard deviation of the feature parameter data set, μ x is the mean of the data set, ν var_out is the variation parameter of the data outside the 95% confidence interval, ν var_in is the variation parameter of the data within the 95% confidence interval, and w1 and w2 are weight coefficients.

[0083] According to the different degrees to which different flame characteristic parameters can reflect the flame combustion stability, corresponding weights are assigned, and the weighted sum of all parameters is taken as the burner combustion flame stability result. The burner flame stability index formula is given as shown below.

[0084]

[0085] In the formula, δ xis the burner flame stability index, m is the number of characteristic parameters used to determine the burner flame stability index, X(j) is the jth characteristic parameter, ω j is the weight corresponding to the jth feature parameter.

[0086] The burner flame stability index evaluates the burner flame combustion stability by comprehensively analyzing the changes in flame characteristic parameters. Since the sum of all weights is set to 1 when determining the weights of different flame characteristic parameters, the range of this index is the same as that of the stability index of a single flame characteristic parameter, and the range is [0,1]. When the value is close to 0, it means that the flame is relatively stable, and when the value is larger, it means that the flame combustion stability is poor.

[0087] Among them, the weight ω of each parameter j , the CRITIC method is used to determine. Assuming there are m samples and n indicators, the evaluation matrix X can be expressed as:

[0088]

[0089] Among them, x i j represents the value of the jth indicator of the ith sample.

[0090] The calculation steps are as follows:

[0091] 1) Index normalization processing:

[0092] Positive indicators:

[0093]

[0094] Contrarian Indicators:

[0095]

[0096] 2) Calculate the mean μ j and standard deviation σ j

[0097]

[0098] where μ j represents the mean of the jth indicator, σ j It represents the comparison strength of the jth indicator. The larger the standard deviation is, the greater the numerical difference of the indicator is, and the more information it can reflect.

[0099] 3) Calculate the coefficient of variation

[0100]

[0101] Among them, ν var_j is the weight corresponding to the jth feature parameter.

[0102] 4) Calculate the correlation coefficient matrix

[0103] ρ ij = cov(y k ,y l ) / (s k ,s j ), k=1, 2,..., n; l=1, 2,..., n;

[0104] Among them, ρ ij is the relationship between the i-th indicator and the j-th indicator, cov(y k ,y l ) represents the covariance between the kth index and the lth index.

[0105] 5) Calculate the information content ηj of the indicator:

[0106]

[0107] Among them, η j Represents the amount of information contained in the j-th indicator and its importance in the overall evaluation system.

[0108] 6) Determine the objective weight as:

[0109]

[0110] where ω j is the weight corresponding to the jth feature parameter.

[0111] Figure 4 The curves of coal feeding amount and primary air volume changing with time; Figure 5 , Figure 6 They are the curves of the unit load and flame stability index changing with time. After analyzing the flame image of the unit under load increase conditions for 18 minutes, it is found that the trend of the flame stability index curve is basically consistent with the trend of the unit load, coal feed rate, and primary air volume curve. When the combustion is under load increase conditions, the flame stability index increases and the combustion stability decreases. It can be seen that the above method has the advantages of fast response speed, high monitoring accuracy, and is not restricted by working conditions, which provides a strong guarantee for the safe and efficient operation of industrial burners.

Claims

1. A burner flame combustion stability monitoring method based on flame image features, characterized in that: The following steps are involved: Step 1: Perform preprocessing operations such as grayscale conversion, filtering and denoising, image enhancement, and flame area segmentation on the flame image collected by the camera in real time; Step 2: Extract flame characteristic parameters; Step 3: Calculate the mean value of the speed magnitude and the mean value of the speed direction of each pixel in the flame image; Step 4: Use the standard deviation to measure the discreteness of the flame image data set; Step 5: Introduce variation parameters to eliminate the dimensionality effect of data; Step 6: Divide the confidence interval of the data set to obtain the stability index of the flame characteristic parameters; Step 7: According to the specific values ​​of the flame image of each frame, the CRITIC method is used to assign corresponding weights to different flame characteristic parameters, and the weighted sum of all flame characteristic parameters is calculated to obtain the burner flame combustion stability.

2. The burner flame combustion stability monitoring method based on flame image features according to claim 1 is characterized in that: The flame characteristic parameters in step 2 include average brightness, unevenness, oscillation frequency, flame temperature, and flame light flow vector velocity magnitude and direction.

3. The burner flame combustion stability monitoring method based on flame image features according to claim 1 is characterized in that: The stability index of the flame characteristic parameters in step 6 is as follows: In the formula, δ par is the flame characteristic parameter stability index, σ x is the standard deviation of the feature parameter data set, μ x is the mean of the data set, ν var_out is the variation parameter of the data outside the 95% confidence interval, ν var_in is the variation parameter of the data within the 95% confidence interval, and w1 and w2 are weight coefficients.

Images