JP2724839B2 - Combustion state diagnosis method and apparatus - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a combustion state diagnosis method and apparatus, and more particularly to a knowledge separation type combustion state diagnosis method and apparatus suitable for a large-scale plant.

[Conventional technology]

Regulation values are set for substances contained in the combustion gas during boiler operation, especially for toxic substances such as nitrogen oxides (hereinafter referred to as NOx). drive.

On the other hand, it is desirable to operate the boiler while always maintaining the combustion efficiency at the maximum. The guideline for calculating the combustion efficiency is the unburned portion in the exhaust gas. As the unburned portion in the exhaust gas increases, the combustion efficiency decreases, and the combustion output increases even with obtaining the same output. However, conventionally, it takes a long time to detect the unburned portion, so that the combustion efficiency during operation has to be relied on experience and intuition.

Recently, coal has been used instead of gas and oil as fuel, and is being converted into pulverized coal, CWM (coal / water slurry), COM (coal / oil slurry), etc. in boilers.

In particular, when coal is used as a fuel, all the nitrogen contained in the fuel itself is converted into NOx by combustion, so that the amount of generated nitrogen is as large as 2000 to 3000 ppm. Further, since the burning rate is extremely slower than that of gas or oil, the unburned content in ash tends to increase with a decrease in furnace temperature, which has been a problem.

On the other hand, as a method of knowing the combustion state during boiler operation,
(1) a flame detector attached to the burner nozzle to detect the flame; (2) a detector attached to the exhaust gas outlet or flue to detect the components contained in the exhaust gas; There was an ITV camera attached to the upper part of the furnace with an explosion-proof mechanism to obtain information.
The detection end shown in FIG. 2 is for detecting (1) the ignition or extinguishing of the burner.
Regarding (2), it is attached to detect whether or not it exceeds a limit value defined by environmental regulations.
The ITV camera (3) is for monitoring the combustion state of the entire furnace.

[Problems to be solved by the invention]

However, such conventionally attached devices have the following disadvantages.

(1) The flame detector is a device that detects “ignition” and “extinguishment” of the flame at the burner outlet. Even though the flame is extinguished, it catches the flame of another burner and mistakes it as “ignition”. Output may be output, and the judgment is left to the operator. This is basically because the frame detector can output only ON (ignition) and OFF (extinguish) signals at the burner outlet.

(2) Exhaust gas component detector requires several tens of seconds to several minutes for analysis, and there is a problem in the real-time property of the analysis value, and the effective value of the component cannot be determined because it is detected at one point. It was only a clue to knowing the state of combustion in the furnace from.

(3) The ITV camera attached to the upper part of the furnace photographs the flame from the opposing burner, but the flame is reflected in a swirling state, so the judgment of the combustion state does not rely on the experience and intuition of the operator. Did not get. Furthermore, when installing an ITV camera, an explosion-proof mechanism was indispensable as a safety measure, and its maintenance was a difficult task.

An object of the present invention is to store, as knowledge, the effect of a model coefficient for estimating a combustion state using a feature amount of a flame image due to a change in combustion conditions, and to sequentially correct the model coefficient using the knowledge to perform combustion. An object of the present invention is to provide a method and an apparatus for estimating a state.

[Means for solving the problem]

The object is to measure the flame as an image, process the image, calculate the feature amount of the flame, and use an estimation model that defines the correlation between the feature amount and the exhaust gas state amount of the flame.
Estimating the exhaust gas state quantity corresponding to the characteristic quantity, and diagnosing the combustion state of the flame based on the estimated quantity, measuring the actual exhaust gas state quantity of the flame and estimating the exhaust gas state quantity The difference between the quantity and the measured quantity is calculated to determine whether the difference is within the limit range, and when the difference exceeds the limit range, the cause of the difference exceeding the limit range and the cause corresponding to the cause are determined. The cause of the difference is inferred in accordance with the knowledge previously describing the correction amount of the coefficient of the estimation model determined in advance, and the estimation model is corrected based on the correction amount of the coefficient corresponding to the inferred cause. Can be achieved.

[Action]

A feature amount is calculated from the image of the flame, and a difference between the estimated exhaust gas state amount of the flame calculated from the feature amount and the pre-stored estimation calculation condition and a measured exhaust gas state amount obtained by measuring the exhaust gas of the flame is calculated. Then, based on the knowledge about the relationship between the difference and the calculation condition of the exhaust gas state quantity stored in advance, the estimation calculation condition is corrected, and the exhaust gas state quantity of the flame is obtained from the corrected estimation calculation condition and the characteristic amount. Is estimated.

〔Example〕

One embodiment of the present invention will be described with reference to FIG. 1, FIG. 3 to FIG.

In the boiler for power generation, as shown in FIG. 1, air is injected into the furnace 12 together with fuel from each stage burner 1, heat generated by burning in the furnace 12 is supplied to the heat transfer tube 13, and exhaust gas is shown in FIG. Not released to the atmosphere from chimneys. In such a boiler, the state of the flame at the base of the burner 1 is
It is empirically known that the state of combustion greatly influences the combustion state including the downstream side.

FIG. 1 is a block diagram showing the entire embodiment. The apparatus of this embodiment measures a flame at the base of a burner 1 as an image using an image fiber 3 protected by a protective tube 2 and installs the image outside the furnace. To the ITV camera 4 The captured flame image is A / D-converted by the A / D converter 5 and stored in the frame memory 6, and subjected to image processing using the processor 7 (7i (i = 1, 2, 3,...)) To perform flame processing. Is extracted. Paying attention to the correlation of the exhaust gas (NOx, unburned portion, etc.) of this feature amount, the exhaust gas component is estimated by an estimation model using the feature amount.

However, exhaust gas components vary greatly depending on load, coal type and fuel type (pulverized coal, CWM (coal / water slurry), etc.). Therefore, by correcting and changing the estimation model using the AI processor 8, a highly accurate estimation result is always obtained.

The flame image taken into the frame memory 6 is subjected to the processing shown in FIG.

The image input processing 7a checks the flame stability and averages it (for example, equation (1)) by judging the presence or absence of a flame, averaging the images, and the like.

Here; average of input images X; instantaneous input images N; averaged sampling number The image preprocessing 7b performs coordinate transformation (rotation of a flame image due to twisting of the image fiber 3, etc., translation of coordinates by clipping an image, etc.). ), Noise removal (removal of noise components included in the flame image), and the like, and the flame image is pre-processed.

Next, the image processing 7c is subjected to half-threshold processing (processing on a grayscale image using the following equation (2)), edge processing, and the like.

Here, (i, j); density at the position of coordinates (i, j) Xth; half-thresholding limit value The feature amount extraction processing 7d performs image processing using area calculation, centroid calculation, perimeter calculation, distance calculation, and the like. Feature values are extracted from.

The exhaust gas estimation processing 7e uses the extracted feature quantity (3)
The exhaust gas state quantity is estimated by the equation. As an example, NOx
(Nitrogen oxide) The estimated amount pNOx is shown.

Where K ₁ , K ₂ , k ₁ , k ₂ , ... Kn; coefficient Flame feature INOX; NOx estimation index A; flame spread B; flame ignitability C; flame temperature (or brightness of each smoke) D) Length of Flame The estimated amount of exhaust gas thus obtained is transmitted to the AI processor 8 through a line.

FIG. 4 shows a flow of the processing of the processor 7 described above.

The transmitted estimated amount of exhaust gas is processed by the AI processor 8 as follows. As an example of the processing, the NOx estimation amount pNOx
Is shown in FIG.

(1) Capture of exhaust gas analysis results CO (carbon monoxide), O ₂ (oxygen), NOx in steady state
(This value is referred to as aNOx).
, And fetched into the AI processor 8 via a PI / O (process input / output device) 11.

(2) Processor that captures the estimation result of processor 7i
7i (i = 1,2,3 ,; number of processors 7)
NOx) is taken into the AI processor 8 through the line.

(3) Judgment of difference between aNOx and pNOx A difference dNOx between aNOx and pNOx is obtained by using equation (4), and it is checked whether or not the difference is within a preset limit range.

dNOx = aNOx-pNOx (4) Whether LL ≦ dNOx ≦ UL where LL; lower limit UL; lower limit dNOx; deviation LL> dNOx, when pNOx exceeds the limit range and the actual NO
A value larger than x is estimated. When dNOx> UL, since the pNOx is estimated to be equal to or smaller than the limit range, the inference unit modifies and changes the estimation model.

(4) Inference (correction / modification of estimation model) The inference unit knows the relationship between the NOx (= aNOx), unburned components, etc., and the flame image during combustion and its features, constants, etc. shown in FIG. Store it in the database.

For example, when dNOx becomes larger than UL, the coal type is changed from the knowledge in Fig. 6 and the fuel ratio of the changed coal type is higher than that of the previous coal type. Calculates the coefficients K ₁ and K ₂ of the estimation model as (K ₁ + ΔK ₁ ), (K ₂
+ ΔK ₂ ) is obtained.

Further, it was found that the burner pattern, load, O ₂ fine particle size, etc. were greatly related to changes in combustion conditions, and the relationship between them and the flame during normal operation was as shown in Table 1 below.

That is, for example, when the O ₂ ratio changes, the equation (5) is obtained.

By storing such a relationship as a knowledge database as in FIG. 6, it is possible to easily add new knowledge and to make more detailed inferences.

(5) Transmission to the processor 7i The conclusion of the inference unit is transmitted to the processor 7i to modify or change the inference model.

(6) Display output The conclusion of the inference unit, and the difference between aNOx and pNOx are displayed. At this time, if the inference process is also displayed, the understanding becomes easier.

(7) End It is determined whether or not all the processors 7i have been completed.

As described above, according to the present embodiment, the combustion state during the boiler operation can be determined satisfactorily by successively correcting or changing the estimation model using knowledge.

Although it is not basically necessary to separate the processors for estimation and inference as in the present embodiment, the load can be reduced because the same knowledge can be used by a plurality of estimation processors, and the knowledge can be assigned to each processor. There is no need to insert it and it can be easily expanded and configured at low cost.

〔The invention's effect〕

According to the present invention, the exhaust gas state is estimated from the flame image, the estimation is compared with the measurement result of the exhaust gas state, and the reason for the difference is determined based on knowledge stored in advance, and the exhaust gas state estimation is performed. Since the combustion state can be accurately grasped by correcting the condition and accurately estimating the exhaust gas state, even an inexperienced operator can efficiently operate a combustion device such as a boiler.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention, FIG. 2 is a conceptual diagram showing a conventional combustion state measuring and monitoring device, FIG. 3 is a block diagram showing internal processing of a processor, and FIG. FIG. 5 is a flowchart showing internal processing, FIG. 5 is a flowchart showing internal processing of the AI processor, and FIG. 6 is a diagram showing an example of knowledge. 1 Burner, 3 Image fiber, 4 ITV camera, 5 A / D converter, 6 Frame memory, 7 Processor, 8 AI processor.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Hiroshi Matsumoto 4026 Kuji-cho, Hitachi City, Ibaraki Prefecture Inside Hitachi, Ltd. Hitachi Research Laboratory, Ltd. (72) Inventor Naganobu 5-2-1 Omika-cho, Hitachi City, Ibaraki Shares (72) Inventor Hisanori Miyagaki 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture In-house Hitachi, Ltd. Omika Plant (72) Inventor Toru Kimura 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture No. Within the Omika Plant of Hitachi, Ltd. (72) Inventor Motoyoshi Sasaki 2310, Kazunuma, Suwayama, Ozai, Kitakanbara-gun, Niigata Prefecture Apartment No. 426 (56) References JP-A-61-93311 (JP, A) JP-A-59-137717 (JP, A)

Claims (2)

(57) [Claims] 1. A method of measuring a flame as an image, processing the image, calculating a characteristic amount of the flame, and using an estimation model in which a correlation between the characteristic amount and an exhaust gas state amount of the flame is defined in advance. ,
Estimating the exhaust gas state quantity corresponding to the characteristic amount, and diagnosing the combustion state of the flame based on the estimated amount, measuring the actual exhaust gas state quantity of the flame, The difference between the estimated amount and the measured amount is calculated to determine whether the difference is within the limit range, and when the difference exceeds the limit range, the cause of the difference exceeding the limit range and the cause According to the knowledge previously describing the correction amount of the coefficient of the estimation model determined in correspondence, the cause of the difference is inferred, and the estimation model is estimated based on the correction amount of the coefficient corresponding to the inferred cause. A method for diagnosing a combustion state, wherein the method is modified. 2. An image capturing apparatus for capturing an image of a flame, an analog-to-digital converter for converting an image captured by the image capturing apparatus from analog to digital, and a frame memory for storing the image converted by the converter. The image stored in the frame memory is subjected to image processing to calculate the feature amount of the flame, and an estimation model is used in which a correlation between the feature amount and the exhaust gas state amount of the flame is defined in advance. A processor for estimating the corresponding exhaust gas state quantity of the flame and diagnosing the combustion state of the flame based on the estimated quantity, wherein the exhaust gas measuring apparatus for measuring the exhaust gas state quantity of the flame And calculating a difference between the estimated amount and the measured amount of the exhaust gas state quantity to determine whether or not the difference is within a limit range. If the difference exceeds the limit range, the difference is set to the limit range. Causes and the The cause of the difference is inferred in accordance with the knowledge previously describing the correction amount of the coefficient of the estimation model determined in accordance with the cause, and the processor is determined based on the correction amount of the coefficient corresponding to the inferred cause. And a knowledge processor for correcting the estimation model.