On-line detection method and device for radiant energy signal in coal-fired boiler

Abstract

The invention is suitable for the technical field of combustion detection of thermal power plants, and provides an online detection method and device for radiant energy signals in a coal-fired boiler, and the method comprises the steps: obtaining a flame image detected by an image probe of the coal-fired boiler; calculating a dirt index of the flame image, and judging whether the image probe is abnormal or not according to the dirt index; if the image probe is abnormal, key operation parameters of the coal-fired power generation unit are obtained; and inputting the key operation parameters into a pre-trained machine learning model to obtain radiation energy signals in the coal-fired boiler. According to the invention, the problem that the radiation energy signal in the furnace cannot be obtained when the probe is severely stained and the probe is cleaned in the existing radiation energy signal detection technology can be solved.

Description

Technical Field

[0001] The invention belongs to the technical field of combustion detection in thermal power plants, and in particular relates to an online detection method and device for radiation energy signals in a coal-fired boiler. Background Art

[0002] At present, coal-fired power generation still occupies a dominant position in power supply, but with the gradual maturity of new energy power generation technology and the continuous increase in installed capacity of new energy, coal-fired power generation units will undertake more deep peak-shaving tasks. A coal-fired power generation unit is a multi-stage energy conversion system, which mainly includes the process from the fuel entering the furnace to the release of radiation energy (fast), the radiation energy is further transferred to the boiler outlet steam (slow) through heat conduction, convection and other methods, and the steam enters the steam turbine and drives the generator to generate electricity (fast), which can be regarded as a process with fast ends and slow in the middle. This means that when existing coal-fired power generation units participate in deep peak-shaving, their load control strategies have no effective way to overcome the impact of the slow process in the middle, resulting in restrictions on the variable load operation capability of the unit and the improvement of the thermal economy of the unit operation.

[0003] The radiation energy signal in the furnace can quickly and accurately reflect the combustion status of the furnace space and has a strong correlation with the unit. When it is introduced into the optimization control of the generator set, it can significantly improve the control quality of the system and is easy to implement in engineering.

[0004] In the related art, the method for detecting the radiation energy signal in the furnace is a radiation imaging-based detection technology for the radiation energy signal of the combustion flame in the furnace of a power plant coal-fired boiler. This technology requires the use of 4-20 high-temperature flame detectors (image probes) installed on the pulverized coal combustion boiler to continuously capture the flame image in the furnace, and extract the combustion radiation energy signal from it for the combustion optimization control of the thermal power unit. This technology requires the opening of observation holes in the furnace of the boiler and the arrangement of image probes at these observation holes. Although fans have been arranged at these measuring points to cool and purge the image probes, when the coal quality changes or the combustion is incomplete, the observation holes are prone to coking and other phenomena, which blocks the normal transmission of the thermal radiation beam and causes the image signal to be distorted by dust and dirt, so that the radiation energy signal in the furnace obtained online will have problems such as unstable signal, large pulsation, and poor reliability. The patent with application number CN201310301791.7 proposes a method for correcting the static deviation of the radiation energy signal. By correcting the radiation energy signal, the static deviation of the radiation energy signal introduced by the factors of the detector dust accumulation, coking, zero drift and the boiler dust accumulation, coking and other operating performance changes is eliminated. However, in boilers burning high-ash coal, the detector lens will have more severe ash contamination problems such as slagging, and in severe cases, it is impossible to obtain measurement results. The observation hole or image probe needs to be cleaned before it can be put into use again. The existing radiation energy signal detection technology cannot obtain the radiation energy signal in the furnace during the cleaning of the probe. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and device for online detection of radiation energy signals in a coal-fired boiler furnace to solve the problem that the existing radiation energy signal detection technology cannot obtain the radiation energy signal in the furnace when the probe is severely contaminated or during the cleaning of the probe.

[0006] A first aspect of an embodiment of the present invention provides an online detection method for radiation energy signals in a coal-fired boiler furnace, comprising:

[0007] Acquire a flame image detected by an image probe of a coal-fired boiler;

[0008] Calculating the dust pollution index of the flame image, and judging whether the image probe is abnormal according to the dust pollution index;

[0009] If the image probe is abnormal, key operating parameters of the coal-fired power generation unit are obtained;

[0010] The key operating parameters are input into a pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

[0011] In combination with the first aspect, in a possible implementation manner of the first aspect, calculating the dust pollution index of the flame image, and judging whether the image probe is abnormal according to the dust pollution index, includes:

[0012] Dividing the flame image into a plurality of regions;

[0013] Under the preset exposure time, determine the brightness value of the pixel points in each area;

[0014] Calculate the number of pixels in each area whose brightness value is lower than a preset brightness threshold;

[0015] according to Calculate the gray dirt index of the flame image; where k j is the dust index, N t ′ is the number of pixels in area t whose brightness value is lower than the brightness threshold, N t is the total number of pixels in region t, and n is the number of divided regions;

[0016] If the dust index is greater than a preset dust index threshold, it is determined that the image probe is abnormal.

[0017] In combination with the first aspect, in a possible implementation manner of the first aspect, after determining whether the image probe is abnormal, the method further includes:

[0018] If the image probe is normal, radiation detection is performed on the flame image to obtain a radiation energy signal in the coal-fired boiler furnace;

[0019] Obtaining key operating parameters of the coal-fired power generation unit;

[0020] The radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit are stored in a preset dedicated database.

[0021] In combination with the first aspect, in a possible implementation of the first aspect, after storing the radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit in a preset dedicated database, the method further includes:

[0022] According to the key operating parameters, the corresponding unit characteristic operating conditions are determined.

[0023] In combination with the first aspect, in a possible implementation of the first aspect, after inputting the key operating parameters into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace, the method further includes:

[0024] Acquiring, from the dedicated database, a radiation energy signal of a unit characteristic operating condition identical to the key operating parameters input into the machine learning model;

[0025] Calculating the statistical average error between the radiation energy signal obtained by the machine learning model and the radiation energy signal of the same unit characteristic operating condition;

[0026] If the statistical average error is less than a preset error threshold, the radiation energy signal and the corresponding key operating parameters obtained by the machine learning model are stored in the dedicated database.

[0027] In combination with the first aspect, in a possible implementation manner of the first aspect, the method further includes:

[0028] Extracting multiple groups of radiation energy signals and corresponding key operating parameters under characteristic operating conditions of different units from the dedicated database;

[0029] Construct a training data set based on multiple groups of radiation energy signals and corresponding unit characteristic operating conditions;

[0030] Based on the training data set, the machine learning model is trained.

[0031] In combination with the first aspect, in a possible implementation of the first aspect, the key operating parameter of the coal-fired power generation unit is at least one;

[0032] Determining the corresponding unit characteristic operating condition according to the key operating parameters includes:

[0033] Performing probability density analysis on each key operating parameter stored in the dedicated database, and dividing the value intervals corresponding to different characteristic operating conditions of each key operating parameter;

[0034] The characteristic operating condition of the unit corresponding to each group of key operating parameters is determined according to the characteristic operating condition corresponding to the numerical range of each key operating parameter in each group of key operating parameters.

[0035] In combination with the first aspect, in a possible implementation of the first aspect, the key operating parameters include at least one of the following: load, total coal quantity, feed water quantity, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening.

[0036] In combination with the first aspect, in a possible implementation of the first aspect, the machine learning model is a convolutional neural network model.

[0037] A second aspect of an embodiment of the present invention provides an online detection device for radiation energy signals in a coal-fired boiler furnace, comprising:

[0038] An acquisition module, used for acquiring a flame image detected by an image probe of a coal-fired boiler;

[0039] A detection module, used to calculate the dust pollution index of the flame image, and determine whether the image probe is abnormal according to the dust pollution index;

[0040] A processing module is used to obtain key operating parameters of the coal-fired power generation unit if the image probe is abnormal; the key operating parameters are input into a pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

[0041] In conjunction with the second aspect, in a possible implementation manner of the second aspect, the detection module is specifically configured to:

[0042] Dividing the flame image into a plurality of regions;

[0043] Under the preset exposure time, determine the brightness value of the pixel points in each area;

[0044] Calculate the number of pixels in each area whose brightness value is lower than a preset brightness threshold;

[0045] according to Calculate the gray dirt index of the flame image; where k j is the dust index, N t ′ is the number of pixels in area t whose brightness value is lower than the brightness threshold, N t is the total number of pixels in region t, and n is the number of divided regions;

[0046] If the dust index is greater than a preset dust index threshold, it is determined that the image probe is abnormal.

[0047] In combination with the second aspect, in a possible implementation manner of the second aspect, after determining whether the image probe is abnormal, the processing module is further used to:

[0048] If the image probe is normal, radiation detection is performed on the flame image to obtain a radiation energy signal in the coal-fired boiler furnace;

[0049] Obtaining key operating parameters of the coal-fired power generation unit;

[0050] The radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit are stored in a preset dedicated database.

[0051] In combination with the second aspect, in a possible implementation of the second aspect, after storing the radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit in a preset dedicated database, the processing module is further used to:

[0052] According to the key operating parameters, the corresponding unit characteristic operating conditions are determined.

[0053] In combination with the second aspect, in a possible implementation manner of the second aspect, after inputting the key operating parameters into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace, the processing module is further used to:

[0054] Acquiring, from the dedicated database, a radiation energy signal of a unit characteristic operating condition identical to the key operating parameters input into the machine learning model;

[0055] Calculating the statistical average error between the radiation energy signal obtained by the machine learning model and the radiation energy signal of the same unit characteristic operating condition;

[0056] If the statistical average error is less than a preset error threshold, the radiation energy signal and the corresponding key operating parameters obtained by the machine learning model are stored in the dedicated database.

[0057] In conjunction with the second aspect, in a possible implementation manner of the second aspect, the processing module is further configured to:

[0058] Extracting multiple groups of radiation energy signals and corresponding key operating parameters under characteristic operating conditions of different units from the dedicated database;

[0059] Construct a training data set based on multiple groups of radiation energy signals and corresponding unit characteristic operating conditions;

[0060] Based on the training data set, the machine learning model is trained.

[0061] In conjunction with the second aspect, in a possible implementation of the second aspect, the key operating parameter of the coal-fired power generation unit is at least one;

[0062] The processing module is also used for:

[0063] Performing probability density analysis on each key operating parameter stored in the dedicated database, and dividing the value intervals corresponding to different characteristic operating conditions of each key operating parameter;

[0064] The characteristic operating condition of the unit corresponding to each group of key operating parameters is determined according to the characteristic operating condition corresponding to the numerical range of each key operating parameter in each group of key operating parameters.

[0065] In combination with the second aspect, in a possible implementation of the second aspect, the key operating parameters include at least one of the following: load, total coal quantity, feed water quantity, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening.

[0066] In combination with the second aspect, in a possible implementation of the second aspect, the machine learning model is a convolutional neural network model.

[0067] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0068] The embodiment of the present invention proposes an online detection method of furnace radiation energy signal based on active training. By establishing a machine learning model to actively learn the detection results of the furnace radiation energy signal, the mathematical relationship between the key operating parameters of the coal-fired power generation unit and the furnace radiation energy signal is obtained. When the image probe cannot work normally due to dust pollution, the established machine learning model is called, and the key operating parameters are input to give the furnace radiation energy signal online, thereby ensuring the continuous effectiveness and reliability of the online detection of the furnace radiation energy signal under harsh measurement environments. In addition, the embodiment of the present invention only needs to compile relevant software programs to achieve it, without adding hardware equipment, and is easy to implement. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0070] Figure 1 The process diagram of the online detection method of the radiation energy signal in the coal-fired boiler provided by the embodiment of the present invention is as follows Figure 1;

[0071] Figure 2 It is a schematic diagram of the structure of the CNN model provided by an embodiment of the present invention;

[0072] Figure 3 The process diagram of the online detection method of the radiation energy signal in the coal-fired boiler provided by the embodiment of the present invention is as follows Figure 2 ;

[0073] Figure 4 It is a structural schematic diagram of an online detection device for radiation energy signals in a coal-fired boiler provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0075] In order to illustrate the technical solution of the present invention, a specific embodiment is provided below for illustration.

[0076] When the fuel burns in the furnace, it releases energy immediately. This process is very fast. For example, the residence time of the pulverized coal in the furnace of a pulverized coal furnace is generally about 2 seconds. A high-temperature flame detector is installed on a pulverized coal combustion boiler to continuously capture the flame image in the furnace. By extracting and calibrating the image pixel information, a combustion radiation energy signal that characterizes the relative size of the radiant heat release in the furnace can be obtained.

[0077] Related technologies:

[0078] The document [Modeling and control simulation of power plant boiler fuel regulation object using radiation energy signal] proposed the idea of ​​using boiler radiation energy to control combustion, and based on the object model of field test, simulated and studied the fuel control strategy with radiation energy signal as the intermediate controlled variable. The results have shown the significance of this method for boiler combustion control. The document [Simulation study of thermal power unit load control system using radiation energy feedback signal] introduced the boiler furnace radiation energy signal as the intermediate controlled variable, and conducted feedback control on the boiler combustion rate, forming a new unit load coordination control method. This control system can improve the tracking performance of the unit load and effectively overcome the influence of the internal disturbance of the boiler combustion rate on the unit operation. The document [Study on boiler oxygen optimization control by introducing radiation energy signal] introduced the furnace radiation energy signal into the oxygen optimization control strategy, which changed the traditional furnace combustion regulation characteristics and improved the control quality of the parameters.

[0079] In summary, the radiation energy signal in the furnace can quickly and accurately reflect the combustion status of the furnace space, and has a strong correlation with the unit. When it is introduced into the optimization control of the generator set, the control quality of the system can be significantly improved, and it is easy to implement in engineering. However, in actual applications, the radiation imaging-based detection technology of the combustion flame radiation energy signal in the furnace of a power plant coal-fired boiler will have dust pollution problems at the image probe lens. In severe cases, the measurement results cannot be obtained at all, which affects the optimization control of the generator set.

[0080] In order to solve this problem, the embodiment of the present invention proposes an online detection method for radiation energy signal in a coal-fired boiler furnace, see Figure 1 As shown, including:

[0081] Step S101, obtaining a flame image detected by an image probe of a coal-fired boiler.

[0082] In this embodiment, there may be more than one image probe. Generally, it is necessary to set up one measuring point at different heights of the coal-fired boiler, and obtain the flame image in the furnace through the image probe installed at each measuring point. Therefore, the flame image obtained here may be one or more.

[0083] Step S102, calculating the dust index of the flame image, and judging whether the image probe is abnormal according to the dust index.

[0084] In this embodiment, by analyzing the dust index of the flame image detected by each image probe, it is determined whether each image probe has a dust abnormality. If any image probe has a dust abnormality, the flame image detected by the image probe is inaccurate, and the radiation energy signal at the corresponding measuring point position of the image probe can no longer be accurately calculated based on the flame image.

[0085] As a possible implementation method, the dirt index can be calculated in the following way.

[0086] Divide the flame image into multiple regions;

[0087] Under the preset exposure time, determine the brightness value of the pixel points in each area;

[0088] Calculate the number of pixels in each area whose brightness value is lower than a preset brightness threshold;

[0089] according to Calculate the gray dirt index of the flame image; where k j is the dust index, N t ′ is the number of pixels in region t whose brightness value is lower than the brightness threshold, N t is the total number of pixels in region t, and n is the number of divided regions;

[0090] If the dust index is greater than a preset dust index threshold, the image probe is determined to be abnormal.

[0091] That is, the flame image obtained by the image probe is divided into n areas, and the exposure time is t e , perform statistical processing on the brightness value R of the pixel points in each area under this exposure time, if the brightness value of the pixel point is lower than the brightness threshold λ R , then the image signal at that location is judged to be gray and distorted, and the gray and dirt index k of each measuring point is calculated by the above formula j (j=1,2,…l). The dirt index k obtained online is j And the set dust index threshold λ k Make a comparison to determine whether the flame image obtained by the image probe is normal. j Less than λ k , the image probe works normally, the flame image obtained is normal, and the calculated radiation energy signal in the furnace is effective and reliable; if k j Greater than or equal to λ k , then there will be a serious dust pollution problem on the image probe, which will cause a large error in the online measurement of the furnace radiation energy signal or the inability to obtain the measurement result.

[0092] Step S103: if the image probe is abnormal, key operating parameters of the coal-fired power generation unit are obtained.

[0093] Here, the key operating parameters may include but are not limited to at least one of the following:

[0094] Load, total coal volume, feed water volume, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening.

[0095] Step S104, inputting key operating parameters into a pre-trained machine learning model to obtain a radiation energy signal in the coal-fired boiler furnace.

[0096] By inputting key operating parameters into the pre-trained machine learning model, the radiation energy signal at each measuring point in the coal-fired boiler furnace can be obtained.

[0097] Here, the machine learning model can be a convolutional neural network (CNN) model, etc. CNN can effectively process multi-dimensional input data, automatically extract local features of input data, and capture complex relationships between different parameters.

[0098] Taking the CNN model as an example, Figure 2The CNN model structure diagram includes data standardization, input layer, convolution layer, pooling layer, flattening layer, fully connected layer, and output layer. The CNN model standardizes the input unit characteristic conditions, reconstructs the pre-processed data into a two-dimensional matrix, performs convolution and pooling operations on it, and then flattens it into a one-dimensional vector and inputs it into the fully connected layer for nonlinear transformation, thereby outputting the radiation energy signal.

[0099] The embodiment of the present invention proposes an online detection method of furnace radiation energy signal based on active training. By establishing a machine learning model to actively learn the detection results of the furnace radiation energy signal, the mathematical relationship between the key operating parameters of the coal-fired power generation unit and the furnace radiation energy signal is obtained. When the image probe cannot work normally due to dust pollution, the established machine learning model is called, and the key operating parameters are input to give the furnace radiation energy signal online, thereby ensuring the continuous effectiveness and reliability of the online detection of the furnace radiation energy signal under harsh measurement environments. In addition, the embodiment of the present invention only needs to compile relevant software programs to achieve it, without adding hardware equipment, and is easy to implement.

[0100] The above embodiment is a process of obtaining the radiation energy signal in the furnace of a coal-fired boiler through a pre-trained machine learning model when the image probe is abnormal. When the image probe is normal, the present invention still detects the radiation energy signal through the image. The radiation energy in the furnace detected when the image probe is working normally and the corresponding key operating parameters of the unit are stored in a dedicated database for training the initial machine learning model, or by updating and repeatedly learning the data, the machine learning model is optimized to continuously adapt to various combustion conditions.

[0101] Combination Figure 3 In one of the following embodiments, a method for online detection of radiation energy signals in a coal-fired boiler furnace is described in more detail, including:

[0102] Step S301, obtaining a flame image detected by an image probe of a coal-fired boiler.

[0103] Step S302, calculating the dust index of the flame image, and judging whether the image probe is abnormal according to the dust index.

[0104] Step S303, if the image probe is normal, perform radiation detection on the flame image to obtain the radiation energy signal in the coal-fired boiler furnace; obtain the key operating parameters of the coal-fired power generation unit; store the radiation energy signal and key operating parameters of the image probe under normal conditions in a preset dedicated database.

[0105] Here, the radiation image processing method is used to obtain the furnace radiation energy signal data RES at l measuring points in the furnace i (i=1,2,…l), and calculate the dust index k of each measuring point j(j=1,2,…l). A special database is established to calculate the dust index k when the image probe is working normally. j , furnace radiation energy signal data RES i And the key operating parameters of the unit P1, P2, ..., P x The data (load, total coal volume, feed water volume, main steam flow, total primary air volume, total secondary air volume) are stored in a dedicated database. The database includes three main data tables: image ash pollution index table, furnace radiation energy signal data table and unit key operating parameter table. Each table contains a unique identifier and a timestamp to facilitate data association and query.

[0106] Step S304, performing probability density analysis on each key operating parameter stored in the dedicated database, dividing the numerical intervals corresponding to different characteristic operating conditions of each key operating parameter; determining the unit characteristic operating conditions corresponding to each group of key operating parameters according to the characteristic operating conditions corresponding to the numerical intervals to which each key operating parameter in each group of key operating parameters belongs.

[0107] In this embodiment, the key operating parameters include load, total coal volume, feed water volume, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening, etc.

[0108] For example, a probability density curve is drawn with load data to determine the load range, and the characteristic operating conditions such as the minimum operating load, the maximum operating load, and the normal operating load and the corresponding numerical range are screened out; then, a probability density analysis is performed on the different total coal quantities under each load to screen out the characteristic operating conditions of different total coal quantities; then, different water supply quantities under each total coal quantity are screened and classified, and so on. By combining the characteristic operating conditions corresponding to each key operating parameter in a group of key operating parameters, the characteristic operating conditions of the unit corresponding to the group of key operating parameters can be determined.

[0109] Step S305, extracting multiple groups of radiation energy signals and corresponding key operating parameters under different unit characteristic conditions from a dedicated database; constructing a training data set based on the multiple groups of radiation energy signals and the corresponding unit characteristic conditions; and training a machine learning model based on the training data set.

[0110] In this embodiment, taking the CNN model as an example, by analyzing the key operating parameters within a certain time range t in the dedicated database, a furnace radiation energy signal RES is constructed. i The key operating parameters are used as input data, and the corresponding radiation energy signals are used as output data to construct a CNN model. During the training process, the CNN model updates the weights of the convolutional layer and the fully connected layer through the back propagation process according to the error between the output radiation energy signal and the target radiation energy signal, and iterates repeatedly to complete the training and learning process of the model, and obtain the mathematical relationship between the key operating parameters and the radiation energy signal.

[0111] Step S306: If the image probe is abnormal, key operating parameters of the coal-fired power generation unit are obtained.

[0112] When the image dirt index k j Greater than the image gray threshold λ k When the image probe cannot work normally, the calculation of the furnace radiation energy signal from the flame image is stopped, and the CNN model established in step S305 is called to calculate the furnace radiation energy signal RES i .

[0113] Step S307, inputting the key operating parameters into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

[0114] Step S308, obtaining from a dedicated database the radiation energy signal of the unit characteristic condition that is the same as the key operating parameters input into the machine learning model; calculating the statistical mean error between the radiation energy signal obtained by the machine learning model and the radiation energy signal of the same unit characteristic condition; if the statistical mean error is less than a preset error threshold, storing the radiation energy signal obtained by the machine learning model and the corresponding key operating parameters in the dedicated database.

[0115] Here, the output furnace radiation energy signal is compared with the furnace radiation energy signal data under the same working conditions in the dedicated database, and the statistical average error σ between the two is given, and the statistical average error is less than the error threshold λ. σ The radiation energy signal in the furnace and the corresponding key operating parameters are stored in a dedicated database. The statistical average error is given by the following formula:

[0116]

[0117] In the formula, RES i is the radiation energy signal at the measuring point i in the furnace output by the convolutional neural network model, is the radiation energy signal at the measuring point i under the same working condition in the dedicated database. When there are multiple similar working conditions in the dedicated database, the average value can be taken for comparison.

[0118] In this embodiment, the calculation results are updated to a dedicated database to ensure data diversity and coverage. By updating the dedicated database, the data is repeatedly learned, the machine learning model is optimized, and the statistical average error of the radiation energy signal in the furnace is reduced to improve its generalization ability, so that the mathematical relationship between key operating parameters and radiation energy signals can continuously adapt to various combustion conditions.

[0119] The embodiment of the present invention addresses the practical problems existing in the online detection technology of the radiation energy signal in the furnace, and proposes an online detection process of the radiation energy in the furnace based on active training. The radiation energy signal in the furnace detected when the image probe is working normally and the corresponding key operating parameters of the unit are stored in a dedicated database, and a machine learning model is established. The key operating parameters are preprocessed and trained to obtain the mathematical relationship between the radiation energy signal and the key operating parameters; when the image probe cannot work normally, the established machine learning model is called, and the key operating parameters are input to give the radiation energy signal in the furnace online; after the image probe is cleaned and starts to work normally, the above process is repeated, and the machine learning model is optimized by updating and repeatedly learning the data to continuously adapt to various combustion conditions. The embodiment of the present invention effectively ensures the continuous effectiveness and reliability of the online detection of the radiation energy signal in the furnace under harsh measurement environments.

[0120] Hereinafter, the method of the present invention is explained by taking a 350MW supercritical boiler as an example.

[0121] (1) Furnace flame image processing:

[0122] The rated power of the boiler is 350MW, and the combustion mode is four-corner tangential combustion. A total of 12 measuring points are set on the front and rear walls of the boiler at heights of 27m, 34m, and 41m, with 4 measuring points set at each height. The image probe at the measuring point is used to obtain the flame image in the furnace, and the radiation image processing method is used to obtain the furnace radiation energy signal data RES at the 12 measuring points in the furnace. i The pixel size of the flame image in the furnace is 706 (width) × 576 (height). The flame image is divided into five regions along the width direction, with pixel ranges of 1-160, 161-300, 301-420, 421-550, and 551-706 respectively; the exposure time t e Set to 1 / 400s, gray threshold λ for 5 regions R They are 50, 70, 90, 110, and 130 respectively, and the image gray threshold λ R is 0.4, and the image dust index k of 12 test points is calculated. j If the value is less than 0.4, the image probe is considered to be working properly and the flame image is normal. If the value is greater than 0.4, the image probe is considered to have a dust pollution problem. The calculated image dust pollution index, furnace radiation energy signal data, and key unit operating parameters are stored in three tables of a dedicated database. Each data item in the table has an auto-incrementing ID as an identifier, and the timestamp records the time when the data is calculated, so as to facilitate data association and query.

[0123] (2) Modeling of the radiation energy training model in the furnace:

[0124] A probability density analysis was performed on the key operating parameters of the units in the dedicated database for the past three months, and a probability density curve of the load was drawn, with the load as the horizontal axis and the frequency as the vertical axis. Three load characteristic operating conditions of 310-330MW, 270-290MW, and 230-250MW were screened out; then a probability density analysis was performed on the total coal quantity under each load, and two total coal quantity characteristic operating conditions of 140-165t / h and 165-190t / h were screened out; and so on, 800-900t / h, 900-1000t / h, and 1 The three characteristic operating conditions of water supply volume of 1000-1100t / h, the three characteristic operating conditions of total primary air volume of 380-400t / h, 400-440t / h, and 440-460t / h, the three characteristic operating conditions of total secondary air volume of 470-610t / h, 610-750t / h, and 750-890t / h are selected, and the calculated radiation energy signal in the furnace corresponding to each set of characteristic operating conditions is found out. A data set including 162 sets of unit characteristic operating conditions and radiation energy signals is constructed. 80% of the data set is taken as the training set, and the remaining 20% ​​of the data set is taken as the test set.

[0125] A convolutional neural network model was established to standardize the key operating parameter data. The preprocessed key operating parameters were reconstructed into a two-dimensional matrix and zero-filled to become a 6×6 matrix. Convolution and pooling operations were performed on it, and then it was flattened into a one-dimensional vector input into the fully connected layer for nonlinear transformation, and then the furnace radiation energy signal data was output. The optimization method used the Adam algorithm with adaptive learning rate, the activation function used the LeakyRelu function optimized on the basis of the Relu function, and the loss function used the mean square error (MSE); in order to prevent the model from overfitting, the Dropout layer was added to the model to improve the generalization ability of the model; 1000 training iterations were performed to optimize the model parameters, so that the training loss function was reduced to 0.1154, and the mathematical relationship between the key operating parameters and the furnace radiation energy signal data was obtained. The test set was used to test the prediction ability of the CNN model. The three load characteristic conditions were predicted and compared with the corresponding target values. After calculation, the average relative errors of the prediction results under the load characteristic conditions of 310-330MW, 270-290MW, and 230-250MW were 2.41%, 2.68%, and 2.64%, respectively.

[0126] (3) Calling the furnace radiation energy training model:

[0127] When the image dirt index k jWhen it is greater than 0.4, that is, the image probe cannot work properly, stop calculating the furnace radiation energy signal from the flame image, call the trained CNN model, input the current load, total coal volume, water supply, main steam flow, total primary air volume, and total secondary air volume of the unit, and then output the furnace radiation energy signal data. Compare the output furnace radiation energy signal with the furnace radiation energy signal data under similar working conditions in the dedicated database, and calculate the statistical average error σ and error threshold λ σ It is set to 0.05, and the furnace radiation energy signal data with a statistical average error less than 0.05 and the corresponding key operating parameters of the unit are stored in a dedicated database.

[0128] (4) Iterative optimization of the furnace radiation energy training model:

[0129] After the image probe is cleaned and the flame image is acquired normally, the image dust index k of the 12 measuring points is calculated. j , and the image gray threshold λ k For comparison, the image dust index, furnace radiation energy signal data and key unit operating parameters when the value is less than 0.4 are updated to the dedicated database, and the training data set is updated at the same time to train the CNN model. Through continuous updating and learning of data, the machine learning model is optimized, the statistical average error of the furnace radiation energy signal is reduced to improve its generalization ability, enhance the stability of the model, and make the mathematical relationship between the key unit operating parameters and the furnace radiation energy signal better adapt to various combustion conditions.

[0130] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0131] Figure 4 4 is a schematic diagram of the structure of an online detection device 40 for radiation energy signals in a coal-fired boiler provided in an embodiment of the present invention. The device comprises:

[0132] An acquisition module 41 is used to acquire a flame image detected by an image probe of a coal-fired boiler;

[0133] The detection module 42 is used to calculate the dust pollution index of the flame image and determine whether the image probe is abnormal according to the dust pollution index;

[0134] The processing module 43 is used to obtain the key operating parameters of the coal-fired power generation unit if the image probe is abnormal; the key operating parameters are input into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

[0135] As a possible implementation manner, the detection module 42 is specifically used for:

[0136] Divide the flame image into multiple regions;

[0137] Under the preset exposure time, determine the brightness value of the pixel points in each area;

[0138] Calculate the number of pixels in each area whose brightness value is lower than a preset brightness threshold;

[0139] according to Calculate the gray dirt index of the flame image; where k j is the dust index, N t ′ is the number of pixels in region t whose brightness value is lower than the brightness threshold, N t is the total number of pixels in region t, and n is the number of divided regions;

[0140] If the dust index is greater than a preset dust index threshold, the image probe is determined to be abnormal.

[0141] As a possible implementation manner, after determining whether the image probe is abnormal, the processing module 43 is further configured to:

[0142] If the image probe is normal, the flame image is subjected to radiation detection to obtain the radiation energy signal in the coal-fired boiler furnace;

[0143] Obtain key operating parameters of coal-fired power generation units;

[0144] The radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit are stored in a preset dedicated database.

[0145] As a possible implementation, after storing the radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit in a preset dedicated database, the processing module 43 is further used to:

[0146] According to the key operating parameters, the corresponding unit characteristic operating conditions are determined.

[0147] As a possible implementation, after the key operating parameters are input into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace, the processing module 43 is further used to:

[0148] Obtaining the radiation energy signal of the unit characteristic condition that is the same as the key operating parameters input into the machine learning model from a dedicated database;

[0149] Calculate the statistical average error between the radiation energy signal obtained by the machine learning model and the radiation energy signal of the same unit characteristic operating condition;

[0150] If the statistical average error is less than the preset error threshold, the radiation energy signal and the corresponding key operating parameters obtained by the machine learning model are stored in a dedicated database.

[0151] As a possible implementation manner, the processing module 43 is further configured to:

[0152] Extract multiple groups of radiation energy signals and corresponding key operating parameters under characteristic operating conditions of different units from a dedicated database;

[0153] Construct a training data set based on multiple groups of radiation energy signals and corresponding unit characteristic operating conditions;

[0154] Train a machine learning model based on the training dataset.

[0155] As a possible implementation, the key operating parameter of the coal-fired power generation unit is at least one;

[0156] The processing module 43 is also used for:

[0157] Conduct probability density analysis on each key operating parameter stored in the dedicated database, and divide the numerical range corresponding to different characteristic working conditions of each key operating parameter;

[0158] The characteristic operating condition of the unit corresponding to each group of key operating parameters is determined according to the characteristic operating condition corresponding to the numerical range of each key operating parameter in each group of key operating parameters.

[0159] As a possible implementation method, the key operating parameters include at least one of the following: load, total coal volume, feed water volume, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening.

[0160] As a possible implementation method, the machine learning model is a convolutional neural network model.

[0161] The embodiment of the present invention proposes an online detection method of furnace radiation energy signal based on active training. By establishing a machine learning model to actively learn the detection results of the furnace radiation energy signal, the mathematical relationship between the key operating parameters of the coal-fired power generation unit and the furnace radiation energy signal is obtained. When the image probe cannot work normally due to dust pollution, the established machine learning model is called, and the key operating parameters are input to give the furnace radiation energy signal online, thereby ensuring the continuous effectiveness and reliability of the online detection of the furnace radiation energy signal under harsh measurement environments. In addition, the embodiment of the present invention only needs to compile relevant software programs to achieve it, without adding hardware equipment, and is easy to implement.

[0162] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0163] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0164] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned embodiments of the online detection method of the radiation energy signal in the furnace of the coal-fired boiler can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium.

[0165] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for online detection of radiation energy signals in a coal-fired boiler, characterized in that: include: Acquire a flame image detected by an image probe of a coal-fired boiler; Calculating the dust pollution index of the flame image, and judging whether the image probe is abnormal according to the dust pollution index; If the image probe is abnormal, key operating parameters of the coal-fired power generation unit are obtained; The key operating parameters are input into a pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

2. The method for online detection of radiation energy signals in a coal-fired boiler according to claim 1, characterized in that: The step of calculating the dust pollution index of the flame image and judging whether the image probe is abnormal according to the dust pollution index includes: Dividing the flame image into a plurality of regions; Under the preset exposure time, determine the brightness value of the pixel points in each area; Calculate the number of pixels in each area whose brightness value is lower than a preset brightness threshold; according to Calculate the gray dirt index of the flame image; where k j is the dust index, N t ′ is the number of pixels in area t whose brightness value is lower than the brightness threshold, N t is the total number of pixels in region t, and n is the number of divided regions; If the dust index is greater than a preset dust index threshold, it is determined that the image probe is abnormal.

3. The method for online detection of radiation energy signals in a coal-fired boiler according to claim 1, characterized in that: After determining whether the image probe is abnormal, the method further includes: If the image probe is normal, radiation detection is performed on the flame image to obtain a radiation energy signal in the coal-fired boiler furnace; Obtaining key operating parameters of the coal-fired power generation unit; The radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit are stored in a preset dedicated database.

4. The method for online detection of radiation energy signals in a coal-fired boiler as claimed in claim 3, characterized in that: After storing the radiation energy signal of the image probe under normal conditions and the key operating parameters of the coal-fired power generation unit in a preset dedicated database, the method further includes: According to the key operating parameters, the corresponding unit characteristic operating conditions are determined.

5. The method for online detection of radiation energy signals in a coal-fired boiler as claimed in claim 4, characterized in that: After inputting the key operating parameters into the pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace, the method further includes: Acquiring, from the dedicated database, a radiation energy signal of a unit characteristic operating condition identical to the key operating parameters input into the machine learning model; Calculating the statistical average error between the radiation energy signal obtained by the machine learning model and the radiation energy signal of the same unit characteristic operating condition; If the statistical average error is less than a preset error threshold, the radiation energy signal and the corresponding key operating parameters obtained by the machine learning model are stored in the dedicated database.

6. The method for online detection of radiation energy signals in a coal-fired boiler as claimed in claim 4, characterized in that: The method further comprises: Extracting multiple groups of radiation energy signals and corresponding key operating parameters under characteristic operating conditions of different units from the dedicated database; Construct a training data set based on multiple groups of radiation energy signals and corresponding unit characteristic operating conditions; Based on the training data set, the machine learning model is trained.

7. The method for online detection of radiation energy signals in a coal-fired boiler as claimed in claim 4, characterized in that: The key operating parameter of the coal-fired power generation unit is at least one; Determining the corresponding unit characteristic operating condition according to the key operating parameters includes: Performing probability density analysis on each key operating parameter stored in the dedicated database, and dividing the value intervals corresponding to different characteristic operating conditions of each key operating parameter; The characteristic operating condition of the unit corresponding to each group of key operating parameters is determined according to the characteristic operating condition corresponding to the numerical range of each key operating parameter in each group of key operating parameters.

8. The method for online detection of radiation energy signals in a coal-fired boiler according to claim 1, characterized in that: The key operating parameters include at least one of the following: load, total coal volume, feed water volume, main steam flow, total primary air volume, total secondary air volume and secondary air regulating damper opening.

9. The method for online detection of radiation energy signals in a coal-fired boiler according to claim 1, characterized in that: The machine learning model is a convolutional neural network model.

10. An online detection device for radiation energy signals in a coal-fired boiler, characterized in that: include: An acquisition module, used for acquiring a flame image detected by an image probe of a coal-fired boiler; A detection module, used to calculate the dust pollution index of the flame image, and determine whether the image probe is abnormal according to the dust pollution index; A processing module is used to obtain key operating parameters of the coal-fired power generation unit if the image probe is abnormal; the key operating parameters are input into a pre-trained machine learning model to obtain the radiation energy signal in the coal-fired boiler furnace.

Images