WO2019225457A1 - Model creation method, plant running assistance method, and model creation device - Google Patents

Abstract

This model creation method is for creating a model indicating the relationship between process values and input parameters for a plant that combusts fuel, the method comprising: a read-in step for reading in plant running data that includes at least physical parameters related to plant specifications or fuel parameters related to the properties of the fuel, and image data for a combustion region of the plant; an extraction step for extracting characteristic quantities of the image data; and a model creation step for creating a model by using at least the physical parameters or the fuel parameters and the characteristic quantities of the image data as input parameters.

Description

Model creation method, plant operation support method, and model creation device

The present disclosure relates to a model creation method, a plant operation support method, and a model creation device.

Patent Document 1 relates to a statistical model (simulation model) for estimating the value of a measurement signal measured when a control signal is applied to a plant, and the statistical model is used to determine the air damper opening, burner angle, and the like at the operation end of the boiler. Is a control device that minimizes each of the process values using the NOx, CO, and H ₂ S concentrations contained in the exhaust gas discharged from the boiler as process values (outputs) of a statistical model. .

Japanese Patent No. 5378288

By the way, when plant specifications and fuel properties change, it is assumed that the relationship between plant input parameters and process values changes. However, Patent Document 1 does not describe any knowledge about how to accurately predict the process value when the plant specification or the fuel property changes.

At least one embodiment of the present invention has been made in view of the conventional problems as described above. The object of the present invention is to accurately predict process values in response to changes in plant specifications and fuel properties. It is to provide a model creation method, a plant operation support method, and a model creation device that can be used.

(1) A model creation method according to at least one embodiment of the present invention includes:
A model creation method for creating a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
A step of reading the operation data of the plant including at least one of physical parameters related to the specifications of the plant or fuel parameters related to the properties of the fuel, and reading image data of the combustion region of the plant;
An extraction step of extracting a feature amount of the image data;
A model creation step of creating the model using the at least one of the physical parameter or the fuel parameter and the feature amount of the image data as the input parameter;
Is provided.

According to the model creation method described in (1) above, the model is created using at least one of the physical parameters or the fuel parameters and the feature amount of the image data as input parameters, so that the difference between the plant specifications and the fuel properties can be reduced. The process value can be predicted with high accuracy using image data in consideration. In addition, since one of the input parameters includes the feature value of the image data of the combustion region, for example, when adopting the plant unburned ash as the process value (the plant input parameter and the plant unburned ash When creating a model that shows the relationship), it is possible to always and quickly evaluate unburned ash in ash, which is usually difficult to measure quickly, reducing fuel cost loss by maintaining optimal combustion conditions can do.

(2) In some embodiments, in the model creation method according to (1) above,
The image data is acquired by photographing the combustion region from above with an imaging device capable of capturing at least one of a moving image or a still image.

According to the model creation method described in (2) above, the process value can be accurately predicted by reading image data taken from above the combustion region and using it for model creation.

(3) In some embodiments, in the model creation method according to (1) or (2) above,
In the extraction step, information relating to any one of the shape, size, color, shade, luminance, temperature (temperature distribution), and wavelength (wavelength distribution) of the combustion region or the amount of change thereof is used as the feature amount of the image data. Extract.

According to the model creation method described in (3) above, the information on the shape, size, color, shade, brightness, temperature (temperature distribution), or wavelength (wavelength distribution) of the combustion region or the amount of change thereof is used. Based on this, the process value can be accurately predicted. In addition, the luminance information of the combustion area is information obtained from the image without detecting the flame itself, for example, even if the flame is not imaged with exhaust gas, for example by installing an imaging device in the upper part of the furnace Information can be obtained. As a result, it is possible to facilitate installation of an imaging device (in-furnace camera) for imaging the inside of the furnace.

(4) In some embodiments, in the model creation method according to any one of (1) to (3) above,
In the model creation step, the model is created using an index related to an exhaust gas component of the plant or an index related to an emission of the plant as the process value.

Exhaust gas components or emissions are generated by the combustion process in the furnace and greatly depend on the combustion state in the combustion region of the plant, so that these generation amounts have a high correlation with the image data of the combustion region. For this reason, as described in (4) above, the process value can be accurately estimated by creating a model using the index related to the exhaust gas component of the plant or the index related to the emission of the plant as the process value.

(5) In some embodiments, in the model creation method according to any one of (1) to (4) above,
In the reading step, a parameter relating to at least one of the structure, performance or design condition of the plant is read as the physical parameter,
In the model creation step, the model is created using at least one of the structure, performance or design conditions of the plant and the feature amount of the image data as the input parameters.

The size of the plant, the installation position of the imaging device, and the fuel injection position differ from plant to plant, and even with the same process value, the feature value varies from plant to plant. For this reason, as described in (5) above, by creating a model in consideration of an appropriate physical parameter related to at least one of the structure, performance, or design conditions of the plant, characteristics resulting from differences in plant specifications It is possible to correct the amount difference and improve the estimation accuracy of the process value.

(6) In some embodiments, in the model creation method according to any one of (1) to (5) above,
In the reading step, a parameter relating to at least one of fuel adjustment, combustion, environmental load or moisture is read as the fuel parameter,
In the model creation step, the model is created using a parameter relating to at least one of fuel adjustment, combustion, environmental load or moisture and the feature quantity of the image data as the input parameters.

Since the process value varies due to changes in the properties of the input fuel, even under the same plant operating conditions, as described in (6) above, at least one of fuel adjustment, combustion, environmental load or moisture By creating a model in consideration of the appropriate fuel parameter according to the above, it is possible to correct the difference in the feature amount due to the difference in the fuel property and improve the estimation accuracy of the process value.

(7) A plant operation support method according to at least one embodiment of the present invention includes:
A plant operation support method using a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
A data reading step for reading operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant;
An extraction step of extracting a feature amount of the image data;
A model reading step for reading the model using the at least one of the physical parameters or the fuel parameters and the feature amount of the image data as the input parameters;
A simulation step of calculating the process value using the operation data of the plant, the image data of the combustion region of the plant, and the model;
An operation instruction step of calculating an operation instruction value of the plant so that the process value satisfies a predetermined condition is further provided.

According to the plant operation support method described in (7) above, by calculating a process value using a model having at least one of a physical parameter or a fuel parameter and a feature amount of image data as input parameters, The process value can be accurately predicted using image data in consideration of the difference between the plant specification and the fuel property. In addition, since one of the input parameters includes the feature value of the image data of the combustion region, for example, when adopting the plant unburned ash as the process value (the plant input parameter and the plant unburned ash When creating a model that shows the relationship), it is possible to always and quickly evaluate unburned ash in ash, which is usually difficult to measure quickly, reducing fuel cost loss by maintaining optimal combustion conditions can do. Thereby, the operation support of a plant can be performed effectively.

(8) A model creation device according to at least one embodiment of the present invention includes:
A model creation device for creating a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
Data acquisition configured to acquire operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant And
A data extraction conversion unit configured to extract a feature amount of the image data acquired by the data acquisition unit;
A model creation unit configured to create the model using the at least one of the physical parameter or the fuel parameter and the feature amount of the image data as the input parameter;
Is provided.

According to the model creation device described in (8) above, by creating a model using at least one of physical parameters or fuel parameters and the feature amount of image data as input parameters, the difference between the plant specifications and the fuel properties can be reduced. The process value can be predicted with high accuracy using image data in consideration.

(9) An operation support apparatus for a plant according to at least one embodiment of the present invention,
A plant operation support apparatus using a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
Data acquisition configured to acquire operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant And
A data extraction conversion unit configured to extract a feature amount of the image data acquired by the data acquisition unit;
A model database that stores the model using the input parameters of at least one of the physical parameter or the fuel parameter and the feature amount of the image data;
A simulation unit that calculates the process value using the operation data of the plant, the image data of the combustion region of the plant, and the model stored in the model database;
An optimization unit configured to determine an optimal set of input parameters such that the process value satisfies a predetermined condition;
An operation instruction unit that calculates an operation instruction value of the plant from the set of the optimal input parameters obtained by the optimization unit;
Is provided.

According to the plant operation support apparatus described in (9) above, by calculating a process value using a model having at least one of a physical parameter or a fuel parameter and a feature amount of image data as input parameters, The process value can be accurately predicted using image data in consideration of the difference between the plant specification and the fuel property. In addition, since one of the input parameters includes the feature value of the image data of the combustion region, for example, when adopting the plant unburned ash as the process value (the plant input parameter and the plant unburned ash When creating a model that shows the relationship), it is possible to always and quickly evaluate unburned ash in ash, which is usually difficult to measure quickly, reducing fuel cost loss by maintaining optimal combustion conditions can do. For this reason, the operation assistance of a plant can be performed effectively.

(10) A plant operation support system according to at least one embodiment of the present invention includes:
A plant operation support system comprising the local support system and a remote support system capable of communicating via a network,
The local support system is:
The driving data, the image data, and the model update result are configured to be transmitted to the remote support system, and the driving data, the image data, and the other local support system transmitted from the remote support system are transmitted. A first transceiver configured to receive model update results;
The remote support system includes:
The driving data, the image data, and the model update result transmitted from each of the local support systems are configured to be received, and the update result is transmitted to all the other local support systems. A second transceiver configured to transmit the image data and the model update result;

According to the plant operation support system described in (10) above, it is possible to share operation data, image data, and model update results in other local support systems (for example, preceding plants).

According to at least one embodiment of the present invention, there is provided a model creation method, a plant operation support method, and a model creation device capable of accurately predicting process values in response to changes in plant specifications and fuel properties. The

1 is a schematic diagram of a boiler plant 100. FIG. It is a model creation flowchart which shows the model creation method which concerns on one Embodiment. It is the flowchart which showed the driving | operation instruction | indication flow which gives the result of the simulation using the model created in FIG. 2 as a driving | operation instruction | indication with respect to an operation control apparatus. 1 is an overall configuration diagram showing a boiler plant 100 and an operation control apparatus 200 that controls the boiler plant 100. FIG. It is a figure which shows an example of the hardware constitutions of the operation control apparatus. It is a schematic diagram showing the relationship between the input and output of a model. It is an example of the data format in model database DB5. It is an example of the data format in plant specification database DB3. It is an example of the data format in fuel property database DB4. It is an example of the data format in additional parameter candidate database DB6. It is a figure which shows the example of whole structure of the driving assistance system 500 which assists model creation. It is a figure which shows the detailed structural example of the driving assistance system 500, and has shown the structure of 300 A of local assistance systems as a representative structural example.

Hereinafter, some embodiments of the present invention will be described with reference to the accompanying drawings. However, the dimensions, materials, shapes, relative arrangements, and the like of the components described in the embodiments or shown in the drawings are not intended to limit the scope of the present invention, but are merely illustrative examples. Absent.
For example, expressions expressing relative or absolute arrangements such as “in a certain direction”, “along a certain direction”, “parallel”, “orthogonal”, “center”, “concentric” or “coaxial” are strictly In addition to such an arrangement, it is also possible to represent a state of relative displacement with an angle or a distance such that tolerance or the same function can be obtained.
For example, an expression indicating that things such as “identical”, “equal”, and “homogeneous” are in an equal state not only represents an exactly equal state, but also has a tolerance or a difference that can provide the same function. It also represents the existing state.
For example, expressions representing shapes such as quadrangular shapes and cylindrical shapes represent not only geometrically strict shapes such as quadrangular shapes and cylindrical shapes, but also irregularities and chamfers as long as the same effects can be obtained. A shape including a part or the like is also expressed.
On the other hand, the expressions “comprising”, “comprising”, “comprising”, “including”, or “having” one constituent element are not exclusive expressions for excluding the existence of the other constituent elements.

Hereinafter, a model creation method for creating a simulation model (hereinafter simply referred to as “model”) indicating a relationship between an input parameter (input) and a process value (output) of a plant that burns fuel will be described.

First, a configuration of a boiler plant 100 is shown as an example of a plant to be modeled by the model creation method. FIG. 1 is a schematic diagram of a boiler plant 100.
The boiler 2 provided in the boiler plant 100 is configured to burn solid fuel. The boiler 2 uses pulverized coal obtained by pulverizing coal as pulverized fuel (solid fuel), combusts the pulverized coal with a combustion burner of the furnace 11, and exchanges heat generated by the combustion with feed water and steam to generate steam. It is a coal fired boiler that can be produced. The fuel is not limited to coal, and may be other fuel that can be burned in a boiler, such as biomass. Further, various kinds of fuels may be mixed and used.

The boiler 2 includes a furnace 11, a combustion device 12, and a flue 13. The furnace 11 has, for example, a hollow shape of a square cylinder and extends along the vertical direction. The wall surface of the furnace 11 is composed of an evaporation tube (heat transfer tube) and a fin connecting the evaporation tube, and the temperature rise of the furnace wall is suppressed by heat exchange with feed water and steam. Specifically, for example, on the side wall surface of the furnace 11, a plurality of evaporation pipes are arranged in the horizontal direction, and each of the evaporation pipes extends along the vertical direction. The fin closes between the evaporation pipe and the evaporation pipe. An inclined surface 62 is provided on the furnace bottom of the furnace 11, and a furnace bottom evaporation pipe 70 is provided on the inclined surface 62 to constitute the bottom surface of the furnace. An image sensor SR1 is provided above the furnace 11.

The combustion device 12 is provided on the lower side in the vertical direction on the furnace wall constituting the furnace 11. Moreover, the combustion apparatus 12 has a plurality of combustion burners (for example, combustion burners 21, 22, 23, 24, 25) mounted on the furnace wall. The plurality of combustion burners 21, 22, 23, 24, 25 are arranged at regular intervals along the circumferential direction of the furnace 11, for example. However, the shape of the furnace, the arrangement of the burners, the number of combustion burners in one stage, and the number of stages are not limited to the above forms.

The combustion burners 21, 22, 23, 24, 25 are respectively supplied to pulverizers (pulverized coal machines / mills) 31, 32, 33, 34, 35 via pulverized coal supply pipes 26, 27, 28, 29, 30. It is connected. Coal is pulverized into a predetermined fine powder size by the pulverizers 31, 32, 33, 34, and 35 when the pulverizers 31, 32, 33, 34, and 35 are fed through a conveyance system (not shown). . The pulverized coal (pulverized coal) is supplied to the combustion burners 21, 22, 23, 24, and 25 from the pulverized coal supply pipes 26, 27, 28, 29, and 30 together with the carrier air (primary air).

Further, the furnace 11 is provided with a wind box 36 at the mounting position of each combustion burner 21, 22, 23, 24, 25. One end of the air duct 37b is connected to the wind box 36, and the other end of the air duct 37b is connected to an air duct 37a that supplies air at a connection point 37d.

Further, a flue 13 is connected vertically above the furnace 11, and a plurality of heat exchangers 41, 42, 43, 44, 45, 46, 47 for generating steam are arranged in the flue 13. ing. The combustion burners 21, 22, 23, 24, 25 inject a mixture of pulverized coal fuel and combustion air into the furnace 11 to form a flame, and combustion gas is generated and flows to the flue 13. Superheated steam is generated by heating the supply water and steam flowing through the furnace wall and heat exchangers 41 to 47 with combustion gas, and a steam turbine (not shown) is rotationally driven by the generated superheated steam and connected to the rotating shaft of the steam turbine. The generator (not shown) is driven to rotate to generate power. An exhaust gas passage 48 is connected to the flue 13. In the exhaust gas passage 48, a denitration device 50 for purifying combustion gas, an air heater 49 for exchanging heat between the air flowing from the forced blower 38 a to the air duct 37 a and the exhaust gas flowing in the exhaust gas passage 48, a dust treatment device 51. , And an induction fan 52 are provided. A chimney 53 is provided at the downstream end of the exhaust gas passage 48. The denitration device 50 may not be provided as long as the exhaust gas standard is satisfied.

Further, the air for conveying pulverized coal (primary air) is sent from the primary air blower 38 b to the air duct 37 e passing through the air heater 49 and the air duct 37 f bypassing the air heater 49. The primary air is merged after the air flow rate of both ducts is adjusted so as to reach a predetermined temperature, etc., and sent to the pulverizers (mills) 31, 32, 33, 34, 35 via the air duct 37g. It is done.

In the furnace 11, after the fuel is excessively combusted by the primary air and the combustion air (secondary air) that is input from the wind box 36 to the furnace 11, the combustion air (after air) is newly input to cause lean fuel combustion. This is a so-called two-stage combustion type furnace. Therefore, the after-air port 39 is provided in the furnace 11. One end of an air duct 37c is connected to the after air port 39, and the other end of the air duct 37c is connected to an air duct 37a for supplying air at a connection point 37d. If the two-stage combustion method is not adopted, the after-air port 39 may not be provided.

The air sent from the blower 38 to the air duct 37a is heated by the air heater 49 by heat exchange with the combustion gas, and is led to the wind box 36 via the air duct 37b and the air duct 37c. Branching to the after air led to the after air port 39 at the connecting point 37d.

The image sensor SR1 is provided on the top of the furnace 11 so as to generate image data of the combustion region in the boiler plant 100. The image sensor SR1 constitutes an imaging device that can capture at least one of a moving image or a still image, for example, and is provided so that the combustion region can be captured from above. The imaging device may be a camera such as a digital camera, a video camera, or an infrared camera capable of detecting a specific wavelength. The image data of the combustion region may be an image itself captured inside the furnace 11, or may be a feature amount that can be extracted from the image. The image may be a still image or a frame (image) that constitutes a moving image. The image sensor SR <b> 1 may be installed at a position other than the upper part of the furnace 11, such as a side wall part of the furnace 11. For example, an image sensor is installed in a portion forming the secondary combustion region in the side wall portion, or a portion further upward (portion closer to the upper portion), and is directed obliquely downward so as to image a lower portion where the primary combustion region is located. May be installed.

FIG. 2 is a model creation flowchart showing a model creation method according to an embodiment.
In the model creation method shown in FIG. 2, a model showing the relationship between the input parameters (inputs) and process values (outputs) of the boiler plant 100 by using image data of the combustion region in the boiler plant 100 that burns fuel. create. The model according to the present embodiment is used for predicting (simulating) a process value based on input parameters input to the model. In principle, a model is created for each process value, but the present invention is not limited to this, and a plurality of process values may be output by one model. Although a boiler plant as a typical power plant will be described here as an example, the model creation method described below is not limited to a boiler plant but can be widely applied to plants that produce industrial products or materials. . For example, as a plant for burning fuel, a steam supply plant and an iron manufacturing plant are exemplified in addition to a power generation plant.

First, in the reading step S1, the operation data of the plant 100 and the image data of the combustion region of the plant 100 are read. The operation data read in the reading step S <b> 1 includes at least one of physical parameters related to plant specifications or fuel parameters related to fuel properties, and process values of the plant 100. When the physical parameter is read in the reading step S1, a parameter related to at least one of the structure, performance, and design conditions of the plant 100 may be read as a physical parameter. Further, the image data of the combustion region of the plant 100 may be either a still image or a moving image that is a continuation of a still image, acquired by the image sensor SR1 installed in the plant 100. The image data format may be TIFF or BMP for still images, uncompressed data represented by AVI for moving images, JPEG or PNG for still images, and MPEG for moving images. Or data in a compression format represented by MP4.

Next, in the extraction step S2, the feature amount of the image data read in the reading step S1 is extracted. The feature amount calculation can be processed in parallel with the reading of the image data by appropriately selecting the configuration of the apparatus, and even when the amount of the image data becomes enormous, the reading of all the image data is completed. The feature amount calculation process can be started from before. The feature amount of the image data relates to, for example, the shape, size, color, shading, brightness, temperature (temperature distribution), or wavelength (wavelength distribution) of the combustion region that appears in the image data or is obtained based on the image data. It may be information or a change amount of at least one of these pieces of information. The amount of change refers to the amount of change in space and / or time. The feature amount is information obtained by quantifying the feature of the image data, which is typically calculated from the luminance information of the image data.

In the case of adopting the above-described change amount or the like as a feature amount, for example, a feature amount obtained by a three-dimensional higher-order local autocorrelation feature amount (CHLAC feature amount), CNN (Convolution Natural Network), or AE (AutoEncoder) is used. Also good. The CHLAC feature amount has an advantage that the spatio-temporal variation appearing in the entire image can be expressed in a compact manner. Further, by extracting feature amounts by convolution processing and pooling processing in CNN and encoding processing in AE, it is possible to effectively reduce and acquire the entire features of the image.

The feature amount extracted based on the luminance information may be a luminance value (luminance value) itself obtained from an image, or may be a statistical value such as an average or peak luminance value. Alternatively, the feature amount may be any of an area, a shape, and a size of a portion having a luminance equal to or higher than a predetermined value, or may be a change amount of the luminance value. Note that the luminance information may include information that can be converted from the luminance value, such as temperature and temperature distribution obtained by converting the luminance value.

The luminance information of the image data is information obtained by converting the shade values of R (red), G (green), and B (blue) in a color image into shade values of black and white (grayscale). It is possible to extract feature amounts of more various patterns by calculating feature amounts from RGB shade information before conversion, that is, image color information. The feature amount extracted based on the color information of the image may be the respective RGB shade values, or may be a statistical value such as an average or a peak shade value. Or the value which shows the correlation between each wavelength like the ratio of the value of R and the value of G may be sufficient, for example.

Luminance information is information obtained from an image without detecting the flame itself. For example, image data can be obtained even when the flame is not imaged with exhaust gas, such as by installing the image sensor SR1 in the upper part of the furnace. Can do. As a result, it is possible to facilitate installation of an imaging device (in-furnace camera) for imaging the inside of the furnace.

Next, in the model creation step S3, the steps shown in detail below are sequentially executed. First, in step S31, 1 is added to the number N of model creations (initial value is 0). In step S32, model creation conditions and additional parameter candidates are read. When the number of times N is 2 or more, the model creation conditions and additional parameter candidates are changed. Here, the model creation conditions are a model creation target (process value), a method (function formula), an allowable error, and the like. Further, the additional parameter candidate is an input parameter addition candidate to be described later.

Next, in step S33, it is determined whether or not the plant specifications (operating conditions) or fuel properties change during operation of the plant 100. If the plant specifications change during operation of the plant 100 in step S33, physical parameters related to the plant specifications are added to the model input parameters in step S34. Here, the physical parameter is a parameter related to at least one of the structure, performance, and design conditions of the plant 100. If the fuel property changes during operation of the plant 100 in step S33, the fuel parameter related to the fuel property is added to the model input parameter in step S35. Here, the fuel parameter is a parameter related to at least one of fuel adjustment, combustion, environmental load, and moisture. If neither the plant specifications nor the fuel properties change during operation of the plant 100 in step S33, the input parameter is not added (step S36).

The size of the plant, the installation position of the imaging device, and the fuel injection position differ from plant to plant, and even with the same process value, the feature value varies from plant to plant. For this reason, as described above, by creating a model in consideration of an appropriate physical parameter related to at least one of the structure, performance, or design conditions of the plant 100, the difference in the feature amount due to the difference in the plant specifications can be reduced. Correction can be made to improve the estimation accuracy of the process value. In addition, the process value varies due to changes in the properties of the fuel that is input, even under the same plant operating conditions. Therefore, as described above, appropriate values related to at least one of fuel adjustment, combustion, environmental load, and moisture By creating a model in consideration of various fuel parameters, it is possible to correct a difference in feature amount due to a difference in fuel properties and improve process value estimation accuracy.

Next, in step S37, the relationship between the input parameters of the model and the process values is machine-learned using the operation data and image data read in the reading step S1, and the model indicating the relationship between the input parameters of the plant 100 and the process values. Create For example, using part of the operation data and part of the image data read in the reading step S1, machine learning is performed on the relationship between the input parameter (physical parameter or fuel parameter and image data) and the process value, and the plant A model showing the relationship between 100 input parameters and process values is created.

Next, in the verification step S4, the steps shown in detail below are sequentially executed. First, in step S41, the accuracy of the model created in step S37 is verified using the operation data and image data not used in machine learning in step S37 among the operation data and image data read in step S1. . For example, the driving data and image data that were not used for machine learning in step S37 out of the driving data and image data read in reading step S1 are input to the model created in step S37. Then, the simulation value (virtual process value) of the process value (output) calculated by the model is compared with the process value (actual process value) as the actual measurement value in the operation data read in the reading step S1. Confirm.

Next, in step S42, it is determined whether or not the error confirmed in step S41 is within the allowable error range. If the error confirmed in step S41 is within the allowable error range, it is determined that the model created in step S37 is valid. If the error confirmed in S41 exceeds the allowable error, it is determined in step S42 that the model created in step S37 is not valid, and it is checked in step S43 whether the number N is equal to or less than the allowable number Nth. . If the number N is equal to or smaller than the allowable number Nth in step S43, the process returns to model creation step S3, and the model is modified by changing the model creation conditions and additional parameter candidates.

Next, in the output step S5, the steps shown in detail below are executed. First, when it is determined in step S42 that the model created in step S37 is valid, in step S51, the model is output to an input / output unit and a model database described later. If the above-mentioned number N exceeds the allowable number Nth in step S43, a model creation error is output in step S52. Either the model or the model creation error is output and the model creation flow ends.

In the output step S5, by outputting the relationship between the model input parameter and the process value, the operator can confirm the relationship (trend) between the input and the output.

According to the model creation method shown in FIG. 2, the model is created by using the physical parameters related to the specifications of the plant 100, the fuel parameters related to the properties of the fuel, and the feature values of the image data of the combustion region as input parameters. The process value can be accurately predicted using image data in consideration of the difference in the plant specifications and the fuel properties. In addition, since one of the input parameters includes the feature value of the image data of the combustion region, for example, when adopting the plant unburned ash as the process value (the plant input parameter and the plant unburned ash When creating a model that shows the relationship), it is possible to always and quickly evaluate the unburned content in ash, which is usually difficult to measure quickly, and reduce the fuel cost and denitration cost by maintaining the optimal combustion state. Loss can be reduced.

2, the method of creating the model of the plant 100 has been described. However, the model created in this way is incorporated into the operation control device 200 (see FIGS. 4A and 4B) of the plant 100 and used. The operation control apparatus 200 of the plant 100 will be described with reference to FIGS. 3, 4A, and 4B.

First, FIG. 3 is a flowchart showing a driving instruction flow for giving a result of a simulation using the model created in FIG. 2 as a driving instruction to the driving control apparatus 200.

First, in step S61, the operation data of the plant 100 and the image data of the combustion region are read (data reading step). In step 62, a model created by the method shown in FIG. 2 is read (model reading step). Further, in step S63, the feature amount of the image data read in step S6 is extracted as in step S2 (extraction step).

Next, simulation is performed in simulation step S7. First, in S71, simulation conditions are set. The simulation condition is a set of input parameters, that is, a set of image data of at least one of a physical parameter related to the specification of the plant 100 or a fuel parameter related to the properties of the fuel and a combustion region of the plant 100.

In step S72, the input parameter set set in step S71 is input to the model created by the method shown in FIG. As a result of the simulation, a process value (virtual process value) of the plant 100 is calculated.

In operation instruction step S8, an operation instruction value for the operation control device 200 is calculated so that the process value satisfies a predetermined condition. First, in step S81, the simulation result is evaluated. In step S82, it is determined whether the virtual process value is optimal (whether a predetermined condition is satisfied). If it is not optimal, the process returns to step S71 to reset the simulation conditions and instruct to calculate a new virtual process value.

The evaluation in step S81 may be performed by converting each virtual process value into a score (dimensionless) with a predetermined conversion coefficient. In step S82, it may be determined that the virtual process value is optimal when the total score value is equal to or greater than a predetermined value. Or, the simulation is performed in multiple cases (simulation conditions), and the virtual process value is optimal when the total value of the scores is the highest among the results, or when the operator determines that the upper case is optimal You may judge. Furthermore, a case with a higher score may be automatically searched using a genetic algorithm or particle swarm optimization technique, and it may be determined whether or not the result is optimal.

Next, in step S83, a driving instruction value is calculated based on the simulation conditions and results determined to be optimal, and the results are output to an output screen or the like to be described later. Thereafter, the driving instruction flow ends.

According to the operation instruction flow, the process value can be accurately estimated by using a general-purpose model, so that the operation support of the plant 100 can be effectively performed.

FIG. 4A is an overall configuration diagram showing the plant 100 and an operation control apparatus 200 that controls the plant 100. FIG. 4B is a diagram illustrating a hardware configuration of the operation control apparatus 200. The operation control device 200 includes a CPU (Central Processing Unit) 72, a RAM (Random Access Memory) 74, a ROM (Read Only Memory) 76, an HDD (Hard Disk Drive) 78, an input I / F 80, and an output I / F 82. These are configured using computers connected to each other via a bus 84. The hardware configuration of the operation control device 200 is not limited to the above, and may be configured by a combination of a control circuit and a storage device. The operation control device 200 is configured by a computer executing a program that realizes each function of the operation control device 200. The function of each part in the operation control apparatus 200 described below is realized, for example, by loading a program held in the ROM 76 into the RAM 74 and executing the program by the CPU 72 and reading and writing data in the RAM 74 and the ROM 76. Further, by providing an arithmetic device specialized in image processing such as GPU (Graphics Processing Unit), it is possible to more efficiently execute image data processing.

The plant 100 includes the configuration shown in FIG. 1 in detail, but FIG. 4A typically shows the image sensor SR1, the physical sensor SR2, and the operation end OP. The operation end OP refers to a valve or a damper.

The configuration of the image sensor SR1 is as described above. The physical sensor SR2 is a sensor for detecting operation data including process values of the plant 100 in addition to a temperature sensor or the like.

The data acquisition unit 301 includes the operation data of the plant 100 (at least one of the physical parameters related to the specifications of the plant 100 or the fuel parameters related to the properties of the fuel, the process value of the plant 100), and the combustion region of the plant 100. And image data. The data acquisition unit 301 acquires image data from the image sensor SR1 of the plant 100 and stores it in the image database DB1. The data acquisition unit 301 acquires operation data from the physical sensor SR2 and stores it in the operation database DB2.

The image database DB1 stores the image data of the fuel region in the plant 100 acquired by the data acquisition unit 301. The operation database DB2 stores the operation data of the plant 100 acquired by the data acquisition unit 301.

The data extraction / conversion unit 302 reads (extracts) data necessary for model creation and operation control from the image database DB1 and the operation database DB2, and performs complementation or format conversion as necessary. Further, an example of the conversion here is a process of estimating and identifying operation data that cannot be directly measured by the image sensor SR1 or the physical sensor SR2 from other data. Since the estimation process is executed in software using a computer, the estimated value is called a soft sensor value. The data extraction / conversion unit 302 reads the image data stored in the image database DB1 and extracts the feature amount of the image data.

The plant specification database DB3 stores physical parameters related to the specifications of the plant 100. A specific example of physical parameters in the plant specification database DB3 will be described later with reference to FIG. The fuel property database DB4 stores fuel parameters related to the properties of fuel used in the boiler plant 100. A specific example of the fuel parameter in the fuel property database DB4 will be described later with reference to FIG.

The model creation unit 303 (model creation device) is configured to create a model using at least one of a physical parameter or a fuel parameter and a feature amount of image data as input parameters. In the illustrated form, the model creation unit 303 uses the operation data and image data from the data extraction / conversion unit 302, the plant specification data from the plant specification database DB3, and the fuel property data from the fuel property database DB4 to generate a boiler. A model indicating the relationship between the input parameters of the plant 100 and the process values is created. The created model is stored and stored in the model database DB5.

The simulation unit 304 calculates a virtual process value using the operation data and the feature amount of the image data output from the data extraction conversion unit 302 and the model output from the model database DB5, and outputs the calculation result to the optimization unit 305. To do.

The optimization unit 305 is configured to calculate an optimal set of input parameters so that the process value satisfies a predetermined condition. The optimization unit 305 determines whether or not the virtual process value is optimal (whether or not the virtual process value satisfies a predetermined condition), and outputs the virtual process value to the operation instruction unit 306 if it is determined to be optimal. If it is determined that it is not, the set of input parameters as simulation conditions is reset and output to the simulation unit 304 so as to perform simulation again. The driving instruction unit 306 calculates a driving instruction value based on the set of input parameters determined to be optimal, and outputs it to the driving control unit 201 (details are described in the driving instruction step).

The input / output unit 307 displays a model creation result and a verification result, a simulation evaluation result, and a driving instruction proposal screen, and accepts an operator's instruction for each. Further, if there is input of additional information to the plant specification database DB3 or the fuel property database DB4, the input result is output to each.

In some embodiments, as illustrated in FIG. 4A, a driving control device 200 including a driving support device 300 and a driving control unit 201 may be configured.
In this case, the operation control unit 201 controls the operation (valve opening degree, etc.) of each operation end OP of the boiler plant 100 based on the operation instruction value output from the operation instruction unit 306 of the operation support device 300. The operation control may be performed automatically based on the operation instruction value, or may be performed after the operator's consent at the input / output unit 307. Further, the operation control unit 201 adds the operation instruction value from the operation support apparatus 300 as a bias value to the operation instruction value from the boiler plant control device (not shown), and instructs the final operation instruction value. May be.

Fig. 5 is a schematic diagram showing the relationship between model inputs and outputs. In addition to the parameters for the operation end OP, the model input parameters include at least one of physical parameters and fuel parameters (both physical parameters and fuel parameters in the illustrated embodiment), and feature amount parameters related to the feature amount of the image data. And are included. The parameter for the operation end OP is a parameter indicating an instruction value (valve opening degree or the like) to the operation end OP. A model is created for each process value. The process value (virtual process value) as an output calculated by simulation using a model includes, for example, unburned ash content, NOx value, CO value, and the like.

Since the amount of exhaust gas components or emissions generated in the plant 100 greatly depends on the combustion state in the combustion region of the plant, the amount of exhaust gas components or emissions generated is highly compatible with image data in the combustion region. For this reason, a process value is obtained by creating a model using a feature amount extracted from image data obtained by imaging the inside of the furnace, using an index related to the exhaust gas component of the plant 100 or an index related to the emission of the plant as a process value. It can be estimated with high accuracy.

In addition, it is desirable that the model is created in consideration of a time lag until the exhaust gas or emission generated in the combustion state indicated by the past image information reaches the state quantity measurement point. Such a time lag may differ so much that it cannot be ignored depending on the type of furnace and the location of the measurement point. By creating a model that takes into account the time lag of image data and process values, an estimation model with high estimation accuracy must be created. Can do.

FIG. 6 is an example of a data format in the model database DB5.
The horizontal axis is segmented by function, input parameter, and model details for each plant. In the function, a technique for modeling is described. Examples of modeling methods include, but are not limited to, stepwise methods, random forests, k-nearest neighbor methods (KNN), and neural network methods. In the model details, each item of the model formula shown below is described. Or the reference destination of another database in which each item is described may be described.
F = f (x, ω, λ, n) (A)
Here, f is a modeling method (function), x is an input parameter, ω is weighted, λ is an intercept, and n is the number of input parameters.

The vertical axis is divided by model creation unit. When creating a model for each process value, process values to be modeled are listed.

FIG. 7 is an example of a data format in the plant specification database DB3. The horizontal axis is divided by plant name.
The vertical axis is divided by items representing plant specifications. Here, a distinction is made between structural specifications and performance specifications. Structural specifications are dimensions, and furnace dimensions are exemplified. The performance specification is a value representing the performance of the plant, and examples thereof include exhaust gas temperature and steam temperature. As the structural specification and performance index, not only the representative value of the measurement result but also the value of the design condition may be described.

FIG. 8 shows an example of a data format in the fuel property database DB4. The horizontal axis is divided by fuel. The vertical axis is divided by items representing fuel properties. The fuel properties include, for example, information on fuel such as fuel components (such as carbon or moisture) and fineness after pulverization, and examples include industrial analysis (such as fuel ratio) and elemental analysis (such as carbon content). Here, the fuel ratio is a ratio of fixed carbon to volatile components.

FIG. 9 is an example of a data format in the additional parameter candidate database DB6.
The horizontal axis is divided according to the data acquisition method and applicable conditions. The data acquisition method may be a method of acquiring a measured value by measurement with the physical sensor SR2, or a method of acquiring a calculated value (soft sensor value) calculated by combining a plurality of measured values. Good. If there is no measured value of an appropriate parameter that represents the plant specification, the calculated value may be substituted.

The vertical axis is divided by physical parameters and fuel parameters. The physical parameters include those obtained from boiler specifications such as structural dimensions and those obtained from measured values or calculated values such as gas temperature and steam temperature. The latter may be set with reference to the design value from the boiler specification.

The fuel parameters are the motor current value of the table of the pulverizer, the pressure acting on the rollers and the differential pressure of the fluid (gas, powder), etc. for coal pulverization, fuel consumption (call flow), heat transfer for coal combustion The amount of heat absorbed by the surface, the boiler output, and the like, the SO ₂ value in the exhaust gas with respect to the environmental load, the inlet air temperature of the pulverizer, etc. with respect to the moisture.

In addition, it is preferable to create a model using operation data of two or more pulverizers as input parameters for the operation data related to the pulverizer. In this case, when calculating the virtual process value, the operation data of one pulverizer is used as a representative, and when the pulverizer stops, the operation data is switched to another pulverizer operation data. This is because even if one pulverizer stops due to maintenance or the like, operation support can be continued using operation data of another pulverizer. What is necessary is just to select a grinder according to the burner position which supplies an operation rate and pulverized coal. In particular, it is preferable to select from a pulverizer that supplies pulverized coal to the middle burner as much as possible. This is because the average behavior in the boiler can be reflected.

FIG. 10 is a diagram illustrating an example of the overall configuration of a driving support system 500 that supports model creation.
The driving support system 500 can communicate with a plurality of local support systems 300A, 300B, and 300C provided for each of the plurality of boiler plants 100A, 100B, and 100C, and the local support systems 300A, 300B, and 300C via the network N, for example. The remote support system 400 is comprised. Boiler plants 100A, 100B, and 100C are connected to local support systems 300A, 300B, and 300C, respectively. Each of the local support systems 300A, 300B, and 300C is connected to the remote support system via the network N.

FIG. 11 is a diagram showing a detailed configuration example of the driving support system 500, and shows the configuration of the local support system 300A as a representative configuration example. The local support systems 300B and 300C have the same configuration as the local support system 300A.

The local support system 300 </ b> A includes a driving support device 300 and a first transmission / reception unit 308. Note that the driving control device 200 may be employed instead of the driving support device 300. The remote support system 400 includes a second transmission / reception unit 401 and a common model database DB7.

The first transmission / reception unit 308 is configured to transmit the operation data, the image data, and the model update result of the plant 100 to the second transmission / reception unit 401 at regular intervals or according to an instruction from the second transmission / reception unit 401. The first transmission / reception unit 308 is configured to receive the operation data, image data, and model update results of the plant 100 in the other local support systems 300B and 300C transmitted from the second transmission / reception unit 401.

The second transmission / reception unit 401 is configured to receive operation data, image data, and model update results transmitted from the local support systems 300A, 300B, and 300C. When a new update result is received, the operation data, the image data, and the model update result are transmitted to the first transmission / reception unit 308 of all other local support systems 300A, 300B, and 300C at any time or at regular intervals. ing.

When the first transmission / reception unit 308 receives new operation data and a model update result, the first transmission / reception unit 308 transmits the operation data update result to the operation database DB2 and the model update result to the model database DB5.

With this configuration, operation data, image data, and model update results in other local support systems 300A, 300B, and 300C can be shared.

The present invention is not limited to the above-described embodiments, and includes forms obtained by modifying the above-described embodiments and forms obtained by appropriately combining these forms.

2 Boiler 11 Furnace 12 Combustion device 13 Flue 21, 22, 23, 24, 25 Combustion burners 26, 27, 28, 29, 30 Pulverized coal supply pipes 31, 32, 33, 34, 35 Crusher 36 Wind box 37a, 37b, 37c, 37e, 37f, 37g Air duct 37d Connection point 38 Blower 38a Pushing blower 38b Primary air blower 39 After air port 41, 42, 43, 44, 45, 46, 47 Heat exchanger 48 Exhaust gas passage 49 Air heater 50 Denitration Device 51 Dust processing device 52 Induction fan 53 Chimney 62 Inclined surface 70 Furnace bottom evaporation pipe 72 CPU
74 RAM
76 ROM
78 HDD
80 input I / F
82 Output I / F
84 Bus 100 Boiler Plant 200 Operation Control Device 201 Operation Control Unit 300 Operation Support Device 300A, 300B, 300C Local Support System 301 Data Acquisition Unit 302 Data Extraction Conversion Unit 303 Model Creation Unit 304 Simulation Unit 305 Optimization Unit 306 Operation Instruction Unit 307 Input / output unit 308 First transmission / reception unit 400 Remote support system 401 Second transmission / reception unit 500 Operation support system DB1 Image database DB2 Operation database DB3 Plant specification database DB4 Fuel property database DB5 Model database DB6 Additional parameter candidate database DB7 Common model database N Network OP Operating end SR1 Image sensor SR2 Physical sensor

Claims (10)

1. A model creation method for creating a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
   A step of reading the operation data of the plant including at least one of physical parameters related to the specifications of the plant or fuel parameters related to the properties of the fuel, and image data of the combustion region of the plant;
   An extraction step of extracting a feature amount of the image data;
   A model creation step of creating the model using the at least one of the physical parameter or the fuel parameter and the feature amount of the image data as the input parameter;
   A model creation method comprising:
2. The model creation method according to claim 1, wherein the image data is acquired by photographing the combustion region from above with an imaging device capable of imaging at least one of a moving image or a still image.
3. The extraction step extracts information relating to any one of the shape, size, color, shading, luminance, temperature, and wavelength of the combustion region or a change amount thereof as the feature amount of the image data. The model creation method described in 1.
4. The model creation according to any one of claims 1 to 3, wherein, in the model creation step, the model is created using an index related to an exhaust gas component of the plant or an index related to an emission of the plant as the process value. Method.
5. In the reading step, a parameter relating to at least one of the structure, performance or design condition of the plant is read as the physical parameter,
   5. The model creation step according to claim 1, wherein in the model creation step, the model is created using the at least one of the structure, performance, or design condition of the plant and the feature amount of the image data as the input parameters. The model creation method described.
6. In the reading step, a parameter relating to at least one of fuel adjustment, combustion, environmental load or moisture is read as the fuel parameter,
   6. The model generation step according to claim 1, wherein in the model creation step, the model is created using a parameter relating to at least one of fuel adjustment, combustion, environmental load or moisture and the feature amount of the image data as the input parameters. The model creation method according to item 1.
7. A plant operation support method using a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
   A data reading step for reading operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant;
   An extraction step of extracting a feature amount of the image data;
   A model reading step for reading the model using the at least one of the physical parameters or the fuel parameters and the feature amount of the image data as the input parameters;
   A simulation step of calculating the process value using the operation data of the plant, image data of a combustion region of the plant, and the model;
   An operation support method for a plant, further comprising an operation instruction step for calculating an operation instruction value for the plant so that the process value satisfies a predetermined condition.
8. A model creation device for creating a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
   Data acquisition configured to acquire operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant And
   A data extraction conversion unit configured to extract a feature amount of the image data acquired by the data acquisition unit;
   A model creation unit configured to create the model using at least one of the physical parameter or the fuel parameter and the feature amount of the image data as the input parameter;
   A model creation device comprising:
9. A plant operation support apparatus using a model indicating a relationship between an input parameter of a plant for burning fuel and a process value,
   Data acquisition configured to acquire operation data of the plant including at least one of a physical parameter related to the specification of the plant or a fuel parameter related to a property of the fuel, and image data of a combustion region of the plant And
   A data extraction conversion unit configured to extract a feature amount of the image data acquired by the data acquisition unit;
   A model database that stores the model using the input parameters of at least one of the physical parameter or the fuel parameter and the feature amount of the image data;
   A simulation unit that calculates the process value using the operation data of the plant, the image data of the combustion region of the plant, and the model stored in the model database;
   An optimization unit configured to determine an optimal set of input parameters such that the process value satisfies a predetermined condition;
   An operation instruction unit that calculates an operation instruction value of the plant from the set of the optimal input parameters obtained by the optimization unit;
   A plant operation support apparatus.
10. A plurality of local support systems including the plant operation support device according to claim 9;
    A plant operation support system comprising the local support system and a remote support system capable of communicating via a network,
    The local support system is:
    The driving data, the image data, and the model update result are configured to be transmitted to the remote support system, and the driving data, the image data, and the other local support system transmitted from the remote support system are transmitted. A first transceiver configured to receive model update results;
    The remote support system includes:
    The driving data, the image data, and the model update result transmitted from each of the local support systems are configured to be received, and the update result is transmitted to all the other local support systems. A plant operation support system including a second transmission / reception unit configured to transmit image data and an update result of the model.

Images