AU2021427367A9 - Operation assistance device, operation assistance method, and operation assistance program - Google Patents

Abstract

Provided are an operation assistance device, an operation assistance method, and an operation assistance program, which are capable of providing effective guidance for a control device to operate equipment by using image information received from a plurality of cameras that monitor the state of each part of the equipment. This operation assistance device acquires image data obtained by photographing the equipment by the cameras, and provides guidance on operation of the equipment by an optimal control algorithm using the image data. The operation assistance device is characterized in that the optimal control algorithm quantifies the image data as features and uses the quantified image data, and the image data is obtained by photographing different positions of the equipment at different time points.

Description

DESCRIPTION TITLE OF INVENTION OPERATION ASSISTANCE DEVICE, OPERATION ASSISTANCE METHOD, AND OPERATION ASSISTANCE PROGRAM TECHNICAL FIELD OF THE INVENTION

[0001]

The present invention relates to an operation

assistance device, an operation assistance method and

an operation assistance program for assisting

manipulation of various types of equipment, plants or

systems (hereinafter referred to simply as

"equipment").

BACKGROUND ART

[0002]

In recent years, in step with technological

innovation such as ICT (Information and Communication

Technology) and IoT (Internet of Thing), the

environments to use a fast calculator, network

communications, a mass data storage device are being

put in place. While attention is focused on the

utilization and application of large amounts of

accumulated data in many industrial fields, a need exists for operation for improving key performance indicators by integrating data collected at field sites, such as measurement data of equipment, data on maintenance checks of equipment, and the like, into a system of managing information on business management and property.

[00031

For example, in the field of power generation

business, due to concerns that the power system

stability is decreased by variation in amount of power

generation with increasing utilization of renewable

energy such as wind power generation, solar power

generation and the like, the importance of thermal

power plants as a backup power source is increased.

The thermal power plants also play not only a role as

load adjustment, but also a role as a base load power

source, and are needed to be operated with

consideration of operational performance such as

efficiency, environmental performance, availability

factor and the like.

[00041

For the purpose of an improvement in operational

performance of such a thermal power plant, Document 1

discloses a control device to cause a reduction in

nitrogen oxide concentrations and/or carbon monoxide concentrations which is environmental performance.

With the technology disclosed in Document 1, a

manipulation signal is generated by combining a model

which simulates the plant characteristics, with

learning means for learning an optimal manipulation

method for the model. Use of the technology allows

the operation conditions to move to optimum values.

Here, the operation conditions are values used in

generation of manipulation signals.

DOCUMENT LIST PATENT DOCUMENT

Document 1: JP 2009-128972

SUMMARY OF INVENTION TECHNICAL PROBLEM

[0006]

In a thermal power plant, various state

quantities (operation data) are measured at different

positions in the thermal power plant, and processing

using the measurement data is performed. Taking a

specific example, in the thermal power plant, if soot

produced when coal burns deposits in a heat exchanger

and/or on a furnace wall, the heat transfer

characteristics change to cause reduction in amount of heat absorption. Also, the soot depositing in the heat exchanger is removed by a soot blower (an injection of steam).

[00071

However, it is difficult to grasp the soot

deposition conditions from only the various state

quantities (operation data) measured at the different

positions in the thermal power plant. Further, there

is a problem of temporarily reducing the efficiency

when, due to a change in heat transfer characteristics

and a decrease of the amount of heat absorption, a

change in heat balance occurs, alternatively, steam is

used by a soot blower.

[00081

On the other hand, for monitoring in various

types of equipment, cameras have conventionally been

used. For example, if the equipment is a thermal

power plant, the combustion state of a burner is

photographed to be monitored. However, in

conventional camera monitoring, the combustion state

is monitored with a focus on time-series changes of

acquired images. Also, different positions of the

thermal power plant are monitored by cameras, but the

monitoring is used to monitor only the photographed

locations, and attention is not focused on the relationship between images photographed at different positions.

[00091

It is obvious that Document 1 or the technology

to use monitoring cameras is applicable to not only

the thermal power plant but also typical equipment.

[0010]

In view of the above, it is an object of the

present invention to provide an operation assistance

device, an operation assistance method and an

operation assistance program which are capable of

providing effective guidance for a control device to

operate equipment by using image information received

from a plurality of cameras that monitor the state of

each part of the equipment.

SOLUTION TO PROBLEM

[0011]

In view of the above, an aspect of the present

invention provides "an operation assistance device

that acquires image data obtained by photographing

equipment with cameras and provides guidance on

operation of the equipment by an optimal control

algorithm using the image data, wherein the optimal

control algorithm quantifies the image data as features and uses the quantified image data, and the image data is obtained by photographing different positions of the equipment at different time points".

[00121

Further, another aspect of the present invention

provides "an operation assistance method for providing

guidance on operation of equipment by an optimal

control algorithm using image data obtained by

photographing the equipment with cameras, wherein the

optimal control algorithm quantifies the image data as

features and uses the quantified image data, and the

image data is obtained by photographing different

positions of the equipment at different time points".

[0013]

Further, another aspect of the present invention

provides "an operation assistance program to provide

guidance on operation of equipment using a plurality

of pieces of image data obtained by photographing a

plurality of positions of the equipment with cameras,

the operation assistance program comprising: a state

recognition program to recognize a state based on one

of the pieces of image data quantified as features; a

state evaluation program to evaluate a state based on

another one of the pieces of image data quantified as

features to obtain an evaluation value; and a learning program to learn action for a transition to a state in which the evaluation value reaches a maximum".

ADVANTAGEOUS EFFECTS OF INVENTION

[0014]

According to the present invention, effective

guidance can be provided for a control device to

operate equipment by using image information received

from a plurality of cameras that monitor the state of

each part of the equipment.

[0015]

For example, if the present invention is applied

to the thermal power plant, the soot deposition

conditions are grasped based on the image data, and

the combustion state may be controlled to prevent the

range of increase in amount of soot deposits from

being extended. State another way, as image

information received from a plurality of cameras that

monitor the state of each part, the image data on the

combustion region and the image data on the soot

deposition conditions are associated with each other

and learned, and therefore the air flow rate balance

can be controlled to achieve the combustion state in

which desired soot deposition conditions occur.

Further, the soot blower is locally operated for a location where soot deposits, so that the need to use the soot blower in unnecessary places is obviated, enabling a reduced amount of steam consumed. By the above, it is possible to provide effective guidance to achieve improved efficiency of the boiler plant.

BRIEF DESCRIPTION OF DRAWINGS

[0016]

Figure 1 is a block diagram illustrating an

example configuration of an operation assistance

device according to an embodiment of the present

invention.

Figure 2 shows flowcharts illustrating the

operation of the operation assistance device according

to an embodiment of the present invention.

Figure 3 is a diagram illustrating an example of

operation data Dl stored in an operation data

database DB11.

Figure 4 is a diagram illustrating an example of

image data D12 stored in an image data database DB2.

Figure 5 is a diagram illustrating an example of

processing operation of preprocessing means 40,

especially, for image data D12.

Figure 6 is a diagram illustrating operation of

the preprocessing means 40 with different data recording cycles.

Figure 7 is a diagram illustrating the result of

extraction of features from image data D12.

Figure 8a is a diagram illustrating a neural

network model as a model included in learning means

70.

Figure 8b is a diagram illustrating the

relationship between input and output of the neural

network model.

Figure 8c is a diagram illustrating an example

result of the learning means 70 being operated.

Figure 9 is a diagram illustrating an example

configuration where a coal fired plant is used as

equipment 19 in Figure 1.

Figure 10a is a diagram illustrating an example

of image data D12a obtained by photographing the

vicinity of a burner 102 as a combustion region of a

boiler 101 with a camera 71a.

Figure 10b is a diagram illustrating an example

of image data D12b obtained by photographing the

vicinity of a heat exchanger 106 of the boiler 101

with a camera 71b.

Figure 11a is a diagram illustrating example of

features extracted by processing the image data D12a

on the combustion region at state recognition means

600.

Figure 1lb is a diagram illustrating example of

features extracted by processing the image data D12b

on the heat exchanger 106 at state evaluation means.

Figure 12 is a flowchart illustrating operational

details of learning means 700 when a coal fired plant

is used as equipment.

Figure 13 is a diagram illustrating a learned

result when a coal fired plant is used as equipment.

DESCRIPTION OF EMBODIMENTS

[0017]

Embodiments according to the present invention

will now be described with reference to the

accompanying drawings.

Embodiment 1

[0018]

Figure 1 is a block diagram illustrating an

example configuration of an operation assistance

device 20 according to an embodiment of the present

invention, and an apparatus associated with it. In

the embodiment, the operation assistance device 20 is

connected to a target apparatus 10 including a control

device 18 and equipment 19 to be controlled, and to an external device 90.

[0019]

The operation assistance device 20 in Figure 1 is

typically configured with a calculator device

(computer). That is, a computing device such as a CPU

executes various processing functions in conformance

with a program for implementing an optimum control

algorithm. The processing functions (optimum control

algorithm) in the computing device may be

schematically shown as including preprocessing means

40, state evaluation means 50, state recognition means

60, learning means 70, action determination means 80

and operation assistance device operation control

means 25 operating these processing functions. It is

noted that each program described in the embodiment

can be delivered to each device via a network or

stored on a recording medium to be distributed. The

processing functions as operation of each section in

the operation assistance device 20 will be detailed in

Figure 2 and later. It is noted that each means

described above can also be implemented as hardware.

It is understood that each element is expressed in the

embodiment using the term "means", but may be

expressed using any term such as "section/portion",

"unit" or the like.

[00201

The operation assistance device 20 includes a

measurement signal database DB1, a processing result

database DB2 as databases DB. The measurement signal

database DB1 includes an operation data database DB11,

an image data database DB12 (DB12a, DB12b, DB12n).

Computerized information is retained in each database

DB, the information being retained in a form typically

designated as an electronic file (electronic data).

It is noted that the databases DB may be located

outside the operation assistance device 20 and

configured to connect via a network.

[00211

The operation assistance device 20 also includes

an external input interface 21 and an external output

interface 22 as interfaces for external connection.

Then, the operation manipulation assistance device 20

is connected to a target apparatus 10 of interest for

application, and an external device 90 through the

interfaces.

[00221

The following aspects are included for

implementation of the operation assistance device 20.

A first one is of the operation assistance device 20

configured as a cloud computing system. This is a configuration in which the operation assistance device

20 is configured on a public network to be available

on the external device 90. A second one is of the

operation assistance device 20 operated and managed by

a company operating the target apparatus 10. This is

a configuration in which the operation assistance

device 20 is connected to an in-house network of an

operational company operating the target apparatus 10,

which are operated and managed by the operational

company in question. The present invention may be

provided by any of the aspects.

[0023]

Then, the external device 90 is configured with a

calculator device (computer). That is, a computing

device such as a CPU executes the following various

processing functions in conformance with a program.

The external device 90 may also be implemented as a

terminal device, such as a tablet, a smartphone, a

note PC and the like. The external device 90 is

provided with an external input device 91 implemented

by a keyboard 92 and a mouse 93, and an image display

device 94. An operator of the target apparatus 10 may

manipulate the external input device 91 based on the

information displayed on the image display device 94

to perform manipulation of the target apparatus 10.

[0024]

The target apparatus 10 also includes a control

device 18 and equipment 19. Here, a measurement

signal Sg70 is transmitted from the equipment 19 to

the control device 18, and a manipulation signal Sg80

is transmitted from the control device 18 to the

equipment 19. The equipment 19 includes a plurality

of cameras 71 (71a, 71b, 71n) installed at different

positions from one another to take images. The

measurement signal Sg70 contains operation data (time

series process value data collected by sensors

installed in the equipment) describing the operation

of the equipment 19, and image data obtained by

photographing with the cameras. Also, the

manipulation signal Sg80 indicates what signal the

control device 18 outputs in response to the

manipulation.

[0025]

The operation assistance device 20 captures an

external input signal Sgl and a measurement signal Sg2

through the external input interface 21, and then the

resulting measurement signal Sg3 is stored in the

measurement signal database DB1. The measurement

signal Sg3 contains operation data Dl and image data

D12 which are respectively stored in an operation data database DB11 and an image data database DB12. It is noted that the image data database DB12 manages the image data D12 at each photographing location. In the embodiment, the image data D12a obtained by photographing with the camera 71a is stored in a location a image database DB12a, the image data D12b obtained by photographing with the camera 71b is stored in a location b image database DB12b, and the image data D12n obtained by photographing with the camera 71n is stored in a location n image database

DB12n.

[0026]

The stored image data DB12 includes images

photographed with a fixed-point camera, a drone, an

underwater drone, a manual camera and/or the like, and

various image data is stored depending on a purpose.

Further, the camera used to photograph may be of

various cameras such as a high-sensitivity CMOS

camera, an infrared camera, a laser camera and the

like, or combinations thereof.

[0027]

The preprocessing means 40 acquires a measurement

signal Sg4 stored in the measurement signal database

DB1, and then applies the data preprocessing as

appropriate before transmitting preprocessed data Sg5 used for state evaluation to the state evaluation means 50 and preprocessed data Sg6 used for state recognition to the state recognition means 60. The preprocessing means 40 performs correction processing with consideration of wasted time and delay time in the equipment on the measurement signal retained in the measurement signal database.

[00281

The state evaluation means 50 extracts features

from the preprocessed data Sg5 used for the state

evaluation to evaluate whether or not the features

have desirable values, and then outputs a state

evaluation result Sg7. The state evaluation result

Sg7 is transmitted to the learning means 70 and the

processing result database DB2.

[0029]

The state recognition means 60 extracts features

from the preprocessed data Sg6 used for the state

recognition to recognize an operation state of the

target apparatus based on the features, and then

outputs a state recognition result Sg8. The state

recognition result Sg8 is transmitted to the learning

means 70, the action determination means 80 and the

processing result database DB2.

[00301

The features extracted by the state evaluation

means 50 and the state recognition means 60 identify a

physical object on the image, and include values into

which the details thereof are coded, size, color,

concentration, temperature, luminance, wavelength, a

rate of change therein, and the like.

[00311

The learning means 70 learns a manipulation

method to allow the state evaluation result Sg7 to

take a desired value, and outputs a learned result

Sg9. The learned result Sg9 is transmitted to the

processing result database DB2. The learned result

Sg9 includes information about action that fits the

current state recognition result. The learning means

70 may be implemented using an optimization algorithm

such as reinforcement learning, a genetic algorithm, a

nonlinear programming method and the like, but the

present invention is not limited to the method of

implementing the learning means 70.

[00321

The processing result database DB2 stores the

state evaluation result Sg7, the state recognition

state Sg8 and the learned result Sg9 which are

obtained as results of operating the state evaluation

means 50, the state recognition means 60 and the learning means 70.

[00331

The action determination means 80 refers to a

learned result Sgl0 and determines an action that fits

the current state recognition result Sg8, and outputs

an action Sgll. The action Sgll is transmitted to the

external output interface 22.

[00341

The external output interface 22 converts the

action Sgll to a recommended manipulation signal Sg12

which is then transmitted to the control device 18 or

the image display device 94. This makes it possible

to use the recommended manipulation signal Sg12 to

control directly the target apparatus 10 or refer to

the recommended manipulation signal Sg12 displayed on

the image display device 94 to manipulate manually the

target apparatus 10.

[00351

It is to be understood that the operation

assistance device 20 according to the embodiment is

illustrated as an example where a computing device

essentially forming a calculator device and databases

DB are installed within the operation manipulation

assistance device 20. However, some of the devices

may be arranged outside the operation assistance device 20 so that only data may be communicated between devices.

[00361

Also, database signals which are signals stored

in each database DB are each displayed through the

external output interface 22 on the image display

device 94. Values of the signals may be also modified

by the external input signal Sgl which is generated

through manipulation of the external input device 91.

[0037]

The external input device 91, which in the

embodiment includes the keyboard 92 and the mouse 93,

may be a device for inputting data, such as a touch

panel, a microphone for voice input and the like.

[00381

It is apparent that the embodiment includes a

method using the operation assistance device 20.

Also, in the embodiment, the guidance target apparatus

10 of interest for application of the operation

assistance device 20 is configured with the control

device 18 and the equipment 19, but it is apparent

that any facility other than having the above

configuration is implementable.

[00391

Figure 2 shows flowcharts illustrating the operation of the operation assistance device 20. The flowcharts are implemented by the operation assistance device operation control means 25 operating each computing device. The operation in the computing device may be divided into two, the function regarding learning and the function regarding actions. The flowchart on the left side of Figure 2 is for using the previously accumulated measurement signals to learn the method of manipulating the target apparatus

10, and the flowchart on the right side of Figure 2 is

for generating the recommended manipulation signal for

the target apparatus 10 based on the learned results.

[00401

Initially, the operation of the function

regarding learning on the left side of Figure 2 is

described. At processing step S10, the previous

measurement signal is captured and stored in the

measurement signal database DB1. At processing step

Sl, the preprocessing means 40 is operated to

generate the preprocessed data Sg5 used for state

evaluation and the preprocessed data Sg6 used for

state recognition from the measurement signal Sg4. At

step Sg 12, the state evaluation means 50 and the

state recognition means 60 are operated to generate

the state evaluation result Sg7 and the state recognition result Sg8. Finally, at processing step

S13, the learning means 70 is operated to generate the

learned result Sg9. The state evaluation result Sg7,

the state recognition result Sg8 and the learned

result Sg9 which are generated in the flowchart are

stored in the processing result database DB2.

[00411

Subsequently, the operation of the function

regarding the action on the right side of Figure 2 is

described. At processing step S20, the latest

measurement signal Sg2 is captured and stored in the

measurement signal database DB1. At processing step

S21, the latest measurement signal Sg4 triggers the

preprocessing means 40 to be operated to generate the

preprocessed data Sg6 used for the state recognition.

At processing step S22, the state recognition means 60

is operated to generate the state recognition result

Sg8. At processing step S23, the action determination

means 80 is operated to generate the action Sgll.

Then, the recommended manipulation signal Sg12 is

transmitted through the external output interface 22

to the control device 18 or the image display device

94.

[00421

At processing step S24, it is determined whether or not operation assistance by the operation assistance device 20 is continuously required. If it is required, the flow returns to processing step S20, but if it is not required, the flow is terminated. It is noted that the methods of determining at processing step S24 whether or not the operation assistance is continuously required include a method by which the operator of the target apparatus 10 uses the external input device 91 to input information about whether or not operation assistance is continuously required, and a determination is made in accordance with the procedure of the information.

[0043]

Figure 3 and Figure 4 are diagrams illustrating

aspects of data stored in the measurement signal

database DB1, in which Figure 3 illustrates an example

of operation data Dl stored in the operation data

database DB11, and Figure 4 illustrates an example of

image data D12 stored in the image data database DB2.

[0044]

As illustrated in Figure 3, for example, time

series data about individual items (item A, item B,

item C, etc.) are stored at sampling periods in the

operation data database DB11. For example, item A is

temperature, item B is a flow rate, and item C is pressure. Also, as illustrated in Figure 4, the distribution of temperatures measured for example on a cross-section of the equipment 19 is stored at sampling periods in the image data database DB2. It is noted that the operation data and the image data on the target apparatus 10 are displayable on the image display device 94.

[0045]

Figure 5 and Figure 6 are diagrams illustrating

examples of the processing operation of the

preprocessing means 40. Initially, Figure 5

especially relates to the image data D12. An overview

of the correction method with an image photographed at

location a and an image photographed at location b is

herein described. Where the time required for a

substance such as fluid or the like to reach location

b from location a (wasted time and delay time) is At12

under the operation conditions 1, the image

photographed at location b is processed after being

corrected in a forward direction by Atl2. Thereby, a

cause-and-effect relation may be learned with

precision. The delay compensation time may be

modified as appropriate while determining operation

conditions, in such a manner as to set time required

for a substance such as fluid or the like to reach location b from location a (wasted time and delay time) to At34 under the operation conditions 2, or the like.

[0046]

In the present invention, images between at least

two points of the upstream and downstream parts of the

fluid or the like flowing in the target equipment are

learned in order to grasp the state of the target

equipment. At this time, rather than comparison

between the images measured at the same time, the

image at the time of measurement at the upstream part

and the image of the fluid in question reaching a

downstream measurement point are required to be

learned. Therefore, in the preprocessing, the delay

time compensation may be performed.

[0047]

Figure 6 is a diagram illustrating the operation

of the preprocessing means 40 with different data

recording cycles. In this connection, there are a

number of variations of the data recording cycle, such

as once a second, once a day, once a week, once a

maintenance cycle (several months or a year) and the

like. For example, where data is recorded once a

second (real time) at location a and recorded once a

week at location n, an average value of features during a period from t5 to t6 at location a is associate with a value of features at t6 at location n.

[00481

The operation in the state evaluation means 50

and the state recognition means 60 will now be

described. In the processing in the means, initially,

for each of a plurality of pieces of image data (D12)

(Dl2a, D12b, D12n) obtained by photographing a

plurality of locations, the features Ci thereof are

extracted from the image data D12, and then, based on

the extracted result, a state of the equipment

photographed is evaluated at the state evaluation

means 50 and the state of equipment photographed is

recognized at the state recognition means 60. The

alphabet "i" as used herein refers to a symbol for

identifying an item of the features. Therefore, the

features Ci include information on a time at which the

image data D12 is acquired, which is time-series

information including all of the features of the image

data D12a, the features of the image data D12b and the

features of the image data D12n, and regarding all of

them, the items of the features are identified by

alphabet "i".

[0049]

Figure 7 illustrates the result of extraction of

features from the image data D12 (Dl2a, D12b, D12n).

The features identify a physical object on the image,

and include values into which the details thereof are

coded, size, color, concentration, temperature,

luminance, wavelength, a rate of change therein, and

the like. These values are extracted as time-series

data. This means that the state of equipment

described by the image data is re-grasped as

quantified features. It is noted that the example of

Figure 7 shows an example of a physical object 1, in

which as long as the plurality of pieces of image data

D12 obtained by photographing a plurality of locations

are about a plurality of physical objects

photographed, a data group in Figure 7 is generated on

a physical object basis.

[00501

In the state evaluation means 50, an evaluation

value E is calculated by use of the following Equation

1, for example.

E = f(Ci) = ZWixCi ... (1)

Here, f(Ci) is a function for calculating an

evaluation value E. In Equation 1, a sum of multiplying a weight parameter Wi and features Ci is defined as an evaluation value. It is noted that the form of regarding f(Ci) can be arbitrarily set depending on its purpose, in addition to the above equation.

[00511

Here, both of the state evaluation means 50 and

the state recognition means 60 extract and use the

features Ci from the image data D12, in which if the

state of one image in the equipment (e.g., the image

data D12a) affects the state of the other image (e.g.,

the image data D12b), the state recognition means 60

may handle the image data D12a on the cause side and

the state evaluation means 50 may handle the image

data D12b on the effect side. Thereby, the state

evaluation means 50 uses the state on the effect side

as an evaluation value, so that the state on the cause

side when the effect is optimized can be clearly

distinguished and grasped.

[00521

The operation of the learning means 70 will now

be described. Features data used for learning in the

learning means 70 is obtained by quantifying the

operation data Dl accumulated in the operation data

database DB11 as well as the image data D12 (Dl2a,

D12b, D12n) which is obtained at a plurality of

locations and accumulated in the image data database

DB12 (DB12a, DB12b, DB12n).

[00531

Embodiment 1 provides a description of the case

of intending to learn an action to minimize the

evaluation value E calculated at the state evaluation

means 50. Figures 8(a) and 8(b) are diagrams

illustrating details of a model included in the

learning means 70. The model is built by use of a

neural network model as shown in Figure 8(a), to

output an evaluation value in response to the input of

the state.

[00541

Figure 8(b) is a diagram illustrating the

relationship between input and output of the neural

network model, in which with the neural network model,

interpolation may be performed between the evaluation

values as an input to determine an evaluation value

for arbitrary state.

[00551

Figure 8c is an example illustrating the result

of operating the learning means 70. In the example in

Figure 8(c), when a current state is in a region A, an

action is determined to increase the state value, and when it is in a region B, an action is determined to decrease the state value. Changing the state in this way causes the evaluation value to take a minimum value so that the evaluation value may be improved.

[00561

It should be understood that in the above

description, because of Figure 8(a), Figure 8(b) and

Figure 8(c), a model incorporated in the learning

means 70 and the learning means 70 may be built by a

neural network model or any other technique.

[00571

According to embodiment 1, because the

information of the camera images photographed at

different positions in the equipment is quantified to

be provided for learning, effective guidance may be

provided for the control device to operate the

equipment by using the image information received from

a plurality of cameras for monitoring a state of each

part of the equipment.

[00581

In further discussion, according to the present

invention, it is possible to acquire, from the image

data, information that is not obtained only from the

operation data measured by the sensors. This is

because the image data is re-grasped as quantified features to be used for learning. In particular, using images photographed at a plurality of locations in different positions is effective at acquiring, for example, a relation in upstream and downstream parts of the fluid, cause and effect events, and the like, through learning.

Embodiment 2

[00591

Embodiment 2 provides a description of an

instance where the operation assistance device and the

operation assistance method which have been described

in embodiment 1 are applied to a coal fired plant.

Figure 9 illustrates an example configuration where a

coal fired plant is used as the equipment 19 in Figure

1. Initially, a mechanism for generating electric

power in the coal fired plant will be briefly

described with reference to Figure 9.

[00601

In Figure 9, a plurality of burners 102 is

installed on the wall surface of a boiler 101

partially forming the coal fired plant as the

equipment 19, the burners 102 supplying: pulverized

coal as fuel into which coal is pulverized at mills

134; primary air for carrying the pulverized coal; and secondary air for adjusting combustion. Then, the pulverized coal supplied through the burners 102 burns within the boiler 101.

[00611

The burners 102 are structured so as to be

arranged to be a plurality of stages in the vertical

direction on the wall surface of the boiler 101 as

illustrated in Figure 9, and a row of a plurality of

burners is arranged in each stage. The structure and

arrangement of the burners illustrated in Figure 9

cause the pulverized coal to burn from the front face

(hereinafter represented as "can's front") and the

rear face (hereinafter represented as "can's back") of

the boiler wall surface within the boiler 101.

Improvement in balance of burner combustion between

the can's front and the can's back offers enhanced

effects of boiler heat recovery and thus improved

thermal efficiency of the plant.

[00621

It is noted that the pulverized coal and the

primary air are guided through piping 139 to the

burners 102 and the secondary air is guided through

piping 141 to the burners 102. The primary air is

guided from a fan 120 to piping 130, which then is

divided midway into piping 132 and piping 131, the piping 132 passing through an air heater 104 located in a downstream part of the boiler 101, the piping 131 bypassing the air heater 104 without passing through.

However, the piping 132 and the piping 131 join

together again to form piping 133 located downstream

of the air heater 104, so that the air is guided into

the mills 134 which are located upstream of the

burners 102 to produce the pulverized coal. The

primary air passing through the air heater 104 is

heated by heat exchange with combustion gas flowing

down the boiler 101. Together with the heated primary

air, the primary air bypassing the air heater 104

carries the coal pulverized at the mills 134, to the

burners 102.

[00631

The mills 134 are arranged (four mills in Figure

9) to correspond to the respective burner stages, and

supply the pulverized coal and the primary air to the

burners forming each stage. That is, if the feed rate

of coal should be reduced due to a reduction in

electric power output, the mill can be stopped to shut

down the burners in each burner stage. In the mill

134, with consideration of the combustion

characteristics of the boiler 101, the number of

revolutions of the mill is adjusted to obtain pulverized coal of a desirable particle size for the properties of coal to be used. The coal stored in a coal bunker 136 is guided to a corresponding coal feeder 135 through a corresponding coal conveyor 137, and is adjusted in feed rate by the coal feeder 135.

Then, the coal is fed into the mill 134 through a coal

conveyor 138.

[00641

Further, the boiler 101 is installed with after

air ports 103 through which two-stage combustion air

is introduced into the boiler 101. The two-stage

combustion air is guided through piping 142 to the

after-air ports 103. In the boiler 101 shown in

Figure 9, the air introduced from piping 140 using a

fan 121 is similarly heated in the air heater 104.

Subsequent to that, the air is divided into the

secondary air piping 141 and the after-air port piping

142, and then guided respectively into the burners 102

and the after-air ports 103 of the boiler 101. The

flow rates of air fed to the burners 102 and the

after-air ports 103 may be adjusted by manipulation of

air dampers (not shown) installed in the respective

piping 141 and piping 142.

[00651

The high-temperature combustion gas, which is generated by burning the pulverized coal within the boiler 101, flows down in the downstream part along a path within the boiler 101 to a heat exchanger 106 located within the boiler 101, which then exchanges heat with feed water in the heat exchanger 106 to generate steam. Subsequently, this results in exhaust gas to flow into the air heater 104 located on a downstream part of the boiler 101, and then the exhaust gas is used in the air heater 104 for heat exchange to raise the temperature of air fed into the boiler 101. As a result, the exhaust gas passing through the air heater 104 is subjected to exhaust gas treatment which is not shown, and then released through a flue into the atmosphere.

[00661

Also, the feed water circulating in the heat

exchanger 106 of the boiler 101 is fed into the heat

exchanger 106 through a feed water pump 105, and then

is overheated in the heat exchanger 106 by the

combustion gas flowing down in the boiler 101, which

is turned into high temperature and pressure steam.

It is understood that although a single heat exchanger

is described in the embodiment, a plurality of heat

exchangers may be installed. Further, the high

temperature and pressure steam generated in the heat exchanger 106 is guided into a steam turbine 108 through a turbine governor 107 so that the steam turbine 108 is driven by the energy of the steam for generation of electric power in a generator 109.

[0067]

Here, the coal fired plant in the embodiment is

installed with various measuring instruments for

detecting state quantities representing the operation

conditions thereof. The operation data Dl, which is a

measurement signal of the coal fired plant acquired

from the measuring instruments installed in the coal

fired plant is stored in the operation database DB11

in the measurement signal database DB1 shown in Figure

1. In the case of the coal fired plant, examples of

sensors (measuring instruments) for acquiring the

operation data Dl includes sensors shown in Figure 9.

Specifically, there are a temperature measuring

instrument 151 for measuring a temperature of the high

temperature and pressure steam fed from the heat

exchanger 106 to the steam turbine 108, a pressure

measuring instrument 152 for measuring a pressure of

the steam, an electric power output measuring

instrument 153 for measuring the electric energy

generated by the generator 109, and the like.

[0068]

Further, the feed water produced by cooling the

steam by a condenser (not shown) of the steam turbine

108 is fed to the heat exchanger 106 of the boiler 101

by the feed water pump 105, and the flow rate of the

feed water is measured by a flow rate measuring

instrument 150. Then, a measurement signal on the

state quantity for a concentration of a component

contained in the exhaust gas which is the combustion

gas discharged from the boiler 101, is measured by a

concentration measuring instrument 154 which is placed

on a downstream part of the boiler 101. It is

apparent that components contained in the exhaust gas

include nitrogen oxides (NOx), carbon monoxide (CO),

hydrogen sulfide (H 2 S), and the like.

[00691

Also, the following are included as measuring

instruments regarding the coal feeder system. A

primary air flowmeter 155 is for measuring the flow

rate of the primary air fed through the piping 133 to

the mill 134, a coal feed meter 156 is for measuring

the coal feed amount of the coal fed to the mill 134

through the coal conveyor 138 by the coal feeder 135,

and a RPM meter 157 is for measuring the number of

revolutions of the mill 134. These are configured to

measure the above information for their corresponding one of the mills and corresponding one of the coal feeders.

[00701

Stated another way, for example, the following

information is used as the operation data Dl

accumulated in the operation data database DB11 in

embodiment 2 which is of application to a coal fired

plant. These are: the coal flow rates fed to the

boiler 101 which is a state quantity of the target

apparatus 10 as the coal fired plant measured by each

of the above measuring instruments; the number of

revolutions of each mill 134 and a first and a second

air flow rate fed to the boiler 101; a feed water

flowrate fed to the heat exchanger 106 of the boiler

101; a temperature of the steam generated at the heat

exchanger 106 of the boiler 101 and fed to the steam

turbine 108; a feed water pressure of the feed water

fed to the heat exchanger 106 of the boiler 101; a gas

temperature and gas concentrations of the exhaust gas

discharged from the boiler 101; an exhaust gas re

circulating flow rate at which a portion of the

exhaust gas discharged from the boiler 101 is re

circulated in the boiler 101; and the like.

[00711

The above information is provided by way of illustration of the operation data Dl stored in the operation database DB11 within the measurement signal database DB1, and additionally, in the present invention, a plurality of sets of image data D12 obtained by photographing a plurality of locations is stored in the image database DB12 within the measurement signal database DB1.

[00721

In embodiment 2 of application to the coal fired

plant, the image data D12 stored in the image data

database DB12 includes, for example: the image data

D12a by a camera 71a photographing the combustion

region on the wall surface of the boiler 101; the

image data D12b by a camera 71a photographing the heat

exchanger 106 of the boiler 101; the image data D12c

obtained by photographing the piping of the boiler

101; and the like. It is noted that a large number of

measuring instruments or cameras, other than the

measuring instruments and cameras illustrated in

Figure 9, are typically installed in the coal fired

plant, which are herein omitted in the figures.

[0073]

Here, points at issue in coal fired power plants

will be clarified. In the coal fired power plant, if

soot produced when coal is burned deposits in the heat exchanger and/or a furnace wall, the heat transfer characteristics change, so that the amount of heat absorption is reduced. Also, the soot depositing in the heat exchanger is removed by a soot blower (an injection of steam). The soot blower is configured to inject steam into each heat exchanger.

[00741

For the problem in question, an analysis is

conventionally performed using the operation data Dl

detected at each position of the thermal power plant

by the sensors, but it is difficult to grasp the soot

deposition conditions from only the operation data

Dl. Further, a change in heat balance occurs due to

a change in heat transfer characteristics and a

reduction in amount of heat absorption. By using

steam by the soot blower, the efficiency is

temporarily reduced.

[0075]

From this, for applying the operation assistance

device 20 according to the present invention to the

coal fired plant, the amount of soot deposits may be

evaluated based on the image photographed when no soot

deposits in the heat exchanger 106 in order to process

the image data D12b about the soot deposits in the

heat exchanger 106.

[0076]

Also, improvement in precision can be achieved by

ensuring consistency between the information grasped

from the image data D12 and the information grasped

from the process data Dl. Specifically, the degree

of fouling estimated from the process data Dl is

corrected based on the soot deposition conditions

grasped from the image data D12b in order to improve

the estimate accuracy of the degree of fouling.

[0077]

The operation assistance device 20 according to

the present invention gasps the soot deposition

conditions based on the image data D12b. And, the

combustion state may be controlled to prevent the

amount of soot deposits from increasing. State

another way, the image data D12a on the combustion

region and the image data D12b on the soot deposition

conditions are associated with each other and learned,

and therefore the air flow rate balance can be

controlled so as to achieve the combustion state in

which desired soot deposition conditions occur.

Further, a location where soot deposits may be grasped

from the image data D12b, so that, by locally blowing

soot, the need to use the soot blower in an

unnecessary place is obviated, enabling a reduced amount of steam consumed. By the above advantageous effects, the efficiency of the boiler plant may be improved.

[00781

An example of the image data D12 when the coal

fired plant is used as equipment will be described

below. Figure 10a is an example of the image data

D12a obtained by photographing the vicinity of the

burners 102 as a combustion region of the boiler 101

with the camera 71a. From the image, the vicinity of

each of the burners 102 arranged in a matrix form on

the wall surface of the boiler 101 exhibits a highest

temperature of 1050 degrees, an outer region around it

exhibits appropriately a temperature of 1000 degrees,

and a further outer region around it exhibits a

temperature of 950 degrees, and a temperature

distribution in the combustion region, such as a

position of each flame and a direction can be grasped.

The temperature distribution is described in the

embodiment, but the image data may include information

on color, concentration, luminance, wavelength and/or

the like.

[0079]

Further, Figure 10b is an example of the image

data D12b obtained by photographing the vicinity of the heat exchanger 106 of the boiler 101 with the camera 71b. From the image, the amount of soot deposits in the heat exchanger 106 may be grasped.

[00801

Attention in this example focuses on the

combustion region located on an upstream part of the

thermic fluid and the heat exchange region located on

a downstream part thereof, and the example has the

cause-and-effect relation between a cause and an

effect in which the upstream state affects the

downstream state. The cameras are arranged in the

above-described positions, so that the combustion

state and the heat exchange state may be photographed,

and only after conversion into quantified features,

they are available for learning, whereby the above

described cause-and-effect relation which has not been

clarified is able to be grasped.

[0081]

An example of the features extracted from the

image data D12 when the coal fired plant is used as

equipment will be described below. Figure 11a is

example of features extracted by processing the image

data D12a on the combustion region at the state

recognition means 60. A mean value of temperatures on

the can's left and the can's right, and the like are extracted as features and stored as time-series data.

[00821

Figure lb is example of features extracted by

processing the image data D12b on the heat exchanger

106 at the state evaluation means 50. The amount of

soot deposits in the heat exchanger 106 such as a

primary superheater (1SH), a secondary superheater

(2SH), and the like are extracted as features and

stored as time-series data.

[00831

Figure 12 is a flowchart illustrating operational

details of the learning means 70 when the coal fired

plant is used as equipment. Initially, at process

step S30, a load plan is captured. The load plan is

herein a plan to output the electric power generation

(load) of the thermal power plant, which is determined

to satisfy the demand for electric power.

[00841

Then, an operation-plan change proposal is

created at process step S31. Here, the operation plan

includes timing to actuate the soot blower, a set

value of the amount of air, a kind of coal (coal

type), and the like.

[00851

At process step S32, the amount of soot deposits when the operation-plan change proposal is performed is estimated. In the process step S32, a state corresponding to the operation plan (a state grasped by the state recognition means 60 when past image data is processed) and the amount of soot deposits are associated with each other and learned, and based on the result, the amount of soot deposits is estimated.

[00861

At process step S33, based on the amount of soot

deposits, efficiency is estimated as a value, and an

evaluation value is calculated. Here, the efficiency

is calculated depending on the amount of soot

deposits. Also, the evaluation value is calculated by

a function of the efficiency as an input, in which the

higher the efficiency, the higher the evaluation

value.

[00871

At process step S34, the quality of the

operation-plan change proposal created in process step

S31 is learned. Specifically, it is learned that an

operation-plan change proposal in which the evaluation

value is increased is a good change proposal and an

operation-plan change proposal in which the evaluation

value is decreased is a bad change proposal, and when

the next operation-plan change proposal is created, a proposal to increase the evaluation value is created.

[00881

At process step S35, learning termination

determination is performed. If YES, the learning is

terminated. If NO, the flow returns to process step

S31. For example, a pre-defined number of times from

process step S30 to process step S35 is repeated, and

the learning is terminated.

[00891

Figure 13 is a diagram illustrating a learned

result when the coal fired plant is used as the

equipment.

In Figure 13, the horizontal axis represents time

and the vertical axis represents, in the order from

top, for example, a load plan in a certain month,

timing of a soot blower injection, a two-stage

combustion ratio which is an example of a set value of

the amount of air, and a coal type.

[00901

An instance is illustrated where the soot blower

injection in Figure 13 is performed on an injection

region that is divided into 4 regions on the can's

right and can's left of the heat exchanger with

respect to the primary heat exchanger 1SH and the

secondary heat exchanger 2SH. For each region, steam is injected into the heat exchanger with appropriate timing to remove the soot depositing in the heat exchanger, leading to an improved amount of heat absorption of the heat exchanger and thus an enhancement in efficiency. On the other hand, since high temperature steam within the plant is used for the soot blower injection, makeup water (pure water etc.) is then required to be added and to be turned into high temperature steam. Because of this, using steam for the soot blower injection causes a temporary reduction in efficiency.

[0091]

Therefore, instead of performing the soot blower

injection on the entire heat exchanger at once, it is

desired that, after division into small regions, the

soot blower injection is performed on each region in a

timely manner. Further, it is desired that the soot

blower injection is directed to a soot deposit

position of the heat exchange. The operation

assistance device 20 according to the present

invention is capable of grasping the soot deposit

position from the image data and directing the soot

blower injection to the position, thus contributing to

an improvement in efficiency.

[0092]

Further, pneumatic operation makes it possible to

maintain a combustion state in which the amount of

soot deposits is decreased, and to choose a coal type

with consideration of soot deposits, thus also

contributing to an improvement in efficiency.

[00931

As described above, using the operation

assistance device 20 according to the present

invention for the coal fired plant enables improved

efficiency of the plant.

Embodiment 3

[00941

In Embodiment 3, a program that should be stored

in ROM of the operation assistance device configured

using a calculator will be described.

[00951

A program to be stored in ROM may be one

corresponding to each processing function in Figure 1,

but main programs herein include a state recognition

program for processing the image data obtained by

photographing a state of the upstream part of the

fluid in the equipment, a state evaluation program for

processing the image data obtained by photographing a

state of the downstream part of the fluid to obtain an evaluation value, and a learning program for using the image processing result for learning to obtain an equipment operation technique as action.

[0096]

The operation assistance program configured with

the above individual programs is utilized as an

appropriately specialized program by the target

apparatus.

LIST OF REFERENCE CHARACTERS

Sgl.. .external input signal,

Sg2...measurement signal,

Sg3...measurement signal,

Sg4...measurement signal,

Sg5.. .preprocessed data to be used for state

evaluation,

Sg6...preprocessed data to be used for state

recognition,

Sg7...state evaluation result,

Sg8...state recognition result,

Sg9.. .learned result,

Sg1O... learned result,

Sgll.. .action,

Sg12.. .recommended manipulation signal,

Sg70...measurement signal,

71...camera,

Sg8O.. .manipulation signal,

10...target apparatus,

18...control device,

19...equipment,

20...operation assistance device,

21...external input interface,

22...external output interface,

DB1...measurement signal database,

DB11.. .operation data database,

DB12.. .image data database,

DB2...processing result database,

40...preprocessing means,

50...state evaluation means,

60...state recognition means,

70...learning means,

80...action determination means,

90...external device,

91...external input device,

92...keyboard,

93...mouse,

94...image display device.

Claims (15)

WHAT IS CLAIMED IS:

1. An operation assistance device that acquires

image data obtained by photographing equipment with

cameras and provides guidance on operation of the

equipment by an optimal control algorithm using the

image data,

wherein the optimal control algorithm quantifies

the image data as features and uses the quantified

image data, and the image data is obtained by

photographing different positions of the equipment at

different time points.

2. The operation assistance device according to

claim 1,

wherein the optimal control algorithm recognizes

a state based on the image data, calculates an

evaluation value based on the image data, and learns

action for a transition to a state in which an

evaluation value reaches a maximum, and

wherein the image data used for the recognition

of the state and the image data used for the

calculation of the evaluation value are obtained by

photographing different positions of the equipment.

3. The operation assistance device according to claim 2, wherein the optimal control algorithm learns using data obtained by preprocessing the image data, and wherein in the preprocessing, image data is corrected with consideration of wasted time and delay time in the equipment.

4. The operation assistance device according to

claim 2 or 3,

wherein the equipment is a boiler plant, and

wherein the operation assistance device acquires

the image data about a combustion portion located on

an upstream part of a thermic fluid in the boiler

plant and about a downstream part of the thermic

fluid, and provides guidance on operation of the

boiler plant such that conditions in the downstream

part of the thermic fluid have desired

characteristics.

5. The operation assistance device according to

claim 4,

wherein the guidance includes parameters of the

boiler or a manipulation method for a soot blower.

6. The operation assistance device according to

claim 4,

wherein the state is recognized using the image

data about the combustion portion on the upstream part

of the thermic fluid in the boiler plant,

wherein the evaluation value is calculated using

the image data about the downstream part of the

thermic fluid, and

wherein guidance defined as action in

reinforcement learning includes parameters of the

boiler or a manipulation method for a soot blower.

7. The operation assistance device according to

claim 6,

wherein when the image data about the downstream

part of the thermic fluid is image data about soot

deposits in a heat exchanger, the amount of soot

deposits is evaluated based on an image of the heat

exchanger without soot deposit on the heat exchanger.

8. An operation assistance method for providing

guidance on operation of equipment by an optimal

control algorithm using image data obtained by

photographing the equipment with cameras,

wherein the optimal control algorithm quantifies the image data as features and uses the quantified image data, and the image data is obtained by photographing different positions of the equipment at different time points.

9. The operation assistance method according to

claim 8,

wherein the optimal control algorithm recognizes

a state based on the image data, calculates an

evaluation value based on the image data, and learns

action for a transition to a state in which an

evaluation value reaches a maximum, and

wherein the image data used for the recognition

of the state and the image data used for the

calculation of the evaluation value are obtained by

photographing different positions of the equipment.

10. The operation assistance method according to

claim 9,

wherein the optimal control algorithm learns

using data obtained by preprocessing the image data,

and

wherein in the preprocessing, image data is

corrected with consideration of wasted time and delay

time in the equipment.

11. The operation assistance method according to

claim 9 or 10,

wherein the equipment is a boiler plant, and

wherein the image data is acquired about a

combustion portion located on an upstream part of a

thermic fluid in the boiler plant and about a

downstream part of the thermic fluid, and guidance on

operation of the boiler plant is provided such that

conditions in the downstream part of the thermic fluid

have desired characteristics.

12. The operation assistance method according to

claim 11,

wherein the guidance includes parameters of the

boiler or a manipulation method for a soot blower.

13. The operation assistance method according to

claim 11,

wherein the state is recognized using the image

data about the combustion portion on the upstream part

of the thermic fluid in the boiler plant,

wherein the evaluation value is calculated using

the image data about the downstream part of the

thermic fluid, and wherein guidance defined as action in reinforcement learning includes parameters of the boiler or a manipulation method for a soot blower.

14. The operation assistance method according to

claim 13,

wherein when the image data about the downstream

part of the thermic fluid is image data about soot

deposits in a heat exchanger, the amount of soot

deposits is evaluated based on an image of the heat

exchanger without soot deposit on the heat exchanger.

15. An operation assistance program to provide

guidance on operation of equipment using a plurality

of pieces of image data obtained by photographing a

plurality of positions of the equipment with cameras,

the operation assistance program comprising:

a state recognition program to recognize a state

based on one of the pieces of image data quantified as

features;

a state evaluation program to evaluate a state

based on another one of the pieces of image data

quantified as features to obtain an evaluation value;

and

a learning program to learn action for a transition to a state in which the evaluation value reaches a maximum.