KR102383774B1 - Method for predicting plant failure and system therefor - Google Patents

Abstract

The present invention relates to a method for predicting plant failure, and to a system for the same. Specifically, from the results of monitoring the operating state of the plant, the normal section and the failure section of the plant are classified, and from the data characteristics of each section It relates to a failure prediction method and a system therefor, characterized in that a prediction data set to be used for predicting a plant failure is generated.

Description

Method for predicting plant failure and a system for the same

The present invention relates to a method for predicting plant failure, and to a system for the same. Specifically, from the results of monitoring the operating state of the plant, the normal section and the failure section of the plant are classified, and from the data characteristics of each section It relates to a failure prediction method and a system therefor, characterized in that a prediction data set to be used for predicting a plant failure is generated.

One plant is provided with a wide variety of installations therein, and operation of the plant is achieved only when all of these installations are operated normally.

On the other hand, due to the characteristics of plants, once a failure occurs, a very large loss occurs, so various methodologies have been devised to predict failure in advance. However, predicting plant failure is a very difficult task for plant operators or plant users. This is an area that is usually not easily accessible to plant operators or plant users, unless they are very skilled professionals.

In other words, the process of analyzing plant data and the process of analyzing characteristic patterns of data is difficult to learn easily from the point of view of non-professional operators or general users because the process is complex and requires advanced knowledge. Even in an environment to utilize machine learning, general operators or general users have difficulties in generating high-quality learning data, and as a result, poor learning data leads to a negative effect on plant failure prediction.

The present invention has been proposed based on the difficulty of predicting plant failures, and the invention to be described through this detailed description can solve the technical problems salpinned above, as well as those of ordinary skill in the art It was invented to provide additional technical elements that cannot be easily invented.

KR 10-1827108 (Registered on 2018.02.01.)

An object of the present invention is to make it possible to easily predict plant failures from a plant operator or user's point of view, and in particular, relatively accurate plant failure prediction is possible even without professional knowledge in analyzing a large amount of plant-related data. aim to make it possible.

In addition, the present invention analyzes the operating state of the plant and generates a series of data sets from a plurality of sections divided into a normal section and a failure section, that is, a prediction data set that can be used to predict plant failure. The purpose.

On the other hand, the technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

The present invention is to solve the above problems, and the method for the system according to the present invention to predict plant failure is (a) receiving general information including at least one of failure detection information, failure diagnosis information, and plant operation information. to do; (b) classifying the operating state of the plant into a plurality of sections with reference to the general information; and (c) generating a prediction data set by analyzing the plurality of sections classified in step (b).

In addition, the method may further include, after step (a), filtering the general information according to a predefined filter condition.

In addition, the method may be characterized in that the plurality of sections are classified on the operating time of the plant.

In addition, in the method, the plurality of sections may be divided into a normal section and a failure section according to the operating state of the plant.

In addition, in the method, the prediction data set may include a normal data set and a failure data set.

In addition, the method may further include; after step (c), the generated prediction data set is provided as learning data for failure prediction modeling.

In addition, in the method, the general information received in step (a) may further include failure prediction information generated by the failure prediction modeling.

On the other hand, a system for predicting a plant failure according to another embodiment of the present invention includes a failure predicting unit, wherein the failure predicting unit provides general information including at least one of failure detection information, failure diagnosis information, and plant operation information. receiving, classifying the operating state of the plant into a plurality of sections with reference to the general information, and analyzing the classified plurality of sections to generate a data set for prediction.

In addition, in the system, the failure prediction unit may filter the general information according to a predefined filter condition.

In addition, the plurality of sections in the system may be characterized in that they are classified on the operating time of the plant.

In addition, in the system, the plurality of sections may be divided into a normal section and a failure section according to the operating state of the plant.

In addition, in the system, the prediction data set may include a normal data set and a failure data set.

In addition, in the system, the failure prediction unit may provide the generated prediction data set as learning data for failure prediction modeling.

In addition, the general information received in the system may further include failure prediction information generated by the failure prediction modeling.

In addition, the system includes a failure detection unit for detecting a failure of the plant; and a failure diagnosis unit that analyzes the failure detected by the failure detection unit, wherein the failure prediction unit receives at least a portion of the information from either the failure detection unit or the failure diagnosis unit can do.

According to the present invention, even an operator or a user without specialized knowledge for data analysis has the effect of being able to predict plant failure, and thus can also achieve the effect of reducing plant maintenance costs.

In addition, according to the present invention, there is an effect that the failure of the plant can be more easily predicted using the data set for prediction.

On the other hand, the effects of the present invention are not limited to those mentioned above, and other technical effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 shows a thermal power plant as a representative example of a plant.
2 shows the overall configuration of the plant system.
3 shows the detailed configuration of a diagnostic system in particular among plant systems.
4 shows a detailed configuration of a failure prediction unit.
5 is a sequence diagram illustrating a process in which data processing for failure prediction is performed by the data processing unit.

The purpose and technical configuration of the present invention, and details regarding the operational effects thereof will be more clearly understood by the following detailed description based on the accompanying drawings in the specification of the present invention. An embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

The embodiments disclosed herein should not be construed or used as limiting the scope of the present invention. It is natural for those skilled in the art that the description including the embodiments of the present specification will have various applications. Accordingly, any embodiments described in the detailed description of the present invention are illustrative for better describing the present invention and are not intended to limit the scope of the present invention to the embodiments.

The functional blocks shown in the drawings and described below are merely examples of possible implementations. Other functional blocks may be used in other implementations without departing from the spirit and scope of the detailed description. Also, although one or more functional blocks of the present invention are represented as separate blocks, one or more of the functional blocks of the present invention may be combinations of various hardware and software configurations that perform the same function.

In addition, the expression that includes certain components is an expression of “open type” and merely refers to the existence of the corresponding components, and should not be construed as excluding additional components.

Furthermore, when it is said that a component is “connected” or “connected” to another component, it may be directly connected or connected to the other component, but it should be understood that other components may exist in between. do.

Hereinafter, a failure prediction system and method proposed by the present invention will be described in detail with reference to the drawings.

First, FIG. 1 shows the general configurations of a thermal power plant as a representative example of a plant. A thermal power plant basically generates steam with thermal power from coal or oil, turns a steam turbine, and produces electrical energy through this. It corresponds to the key composition of boiling. The main components of a boiler include a boiler body containing water and steam, a fuel combustion device, and a furnace. Of these, the combustion device and furnace are controlled by the control system to control temperature and pressure.

On the other hand, as seen in the schematic configuration of the thermal power plant above, one plant produces electrical energy through a complex process. presupposes to be The problem is when one of many parts malfunctions, or when an algorithm malfunctions and the entire plant stops working due to a failure. There is a problem in that it is not easy to determine whether or not it has been done, so it takes a considerable amount of money and time to restore it.

The present invention identifies the data characteristics of the section in which the problem occurs and the section in which the plant operates normally by constantly monitoring the failure of the plant, and generates a data set that can predict the failure of the plant based on the identified data characteristics, thereby preventing the failure of the plant. It has an important purpose to predict in advance and prepare for it.

Hereinafter, with reference to the drawings, a method for predicting plant failure proposed in the present invention and a system for the same will be described.

2 shows an overall overview of a plant system in which a plant failure prediction method according to the present invention is executed, and the plant system referred to in this detailed description is premised on a minimum environment in which control can be made in a general plant, As shown in FIG. 2 , the control system 10 , the power generation system 20 , the diagnostic system 30 , and the adjustment system 40 may be further included. For reference, in this detailed description, each system may be implemented by hardware each having a central processing unit and a memory, or all systems may be implemented by one hardware (one having a central processing unit and a memory). . In addition, it should be understood that software for implementing each system may be implemented on a program designed in a computer-readable language and executed by the central processing unit (CPU). Furthermore, each system is implemented by hardware, firmware, software, or a combination thereof, and when implemented using hardware, an ASIC (application specific integrated circuit) or DSP (digital signal processor), DSPD (digital signal processing device), programmable logic device (PLD), field programmable gate array (FPGA), etc., if implemented using firmware or software, includes modules, procedures or functions that perform the above functions or operations It can be configured as firmware or software to

The control system 10 is a configuration for controlling all operations of the plant, and may largely function to grasp the operating state of the plant and to control the operation of the plant according to a predetermined algorithm. The control system 10 may receive an operation input of a user (or an operator) input from the outside, and may transmit a control command to the power generation system 20 based on the operation input, and in this process, the adjustment system 40 After receiving the operation result of , it may be reflected in the control command.

The power generation system 20 is configured to operate the actual plant after receiving a control command from the control system 10. In the power generation system 20, a plurality of actuators, plant dynamics, and Sensors may be included. `

The diagnostic system 30 is a configuration that detects and analyzes plant failures and further predicts failures that may occur in the future based on the above detection and analysis. It will be discussed, so we will postpone the detailed description here.

Finally, the reconfiguration system 40 is configured to receive the analysis or prediction operation result from the diagnosis system 30 , and provide feedback information to the control system 10 based on the result of the analysis or prediction operation. The control system 40 provides feedback information to the plant in a direction to reduce the possibility of failure, and there is no limitation on the type of feedback information provided as long as it is for this purpose. However, preferably, the adjustment system 40 may provide feedback information to prevent failure in advance by limiting various power generation parameters to the extent that failure does not occur, and the feedback information provided at this time is the control system 10 ) may include feedback information for instructing to operate at a value lower than the power generation parameter in the control command transmitted to the power generation system 20 in the previous cycle.

3 is a more detailed view of the diagnostic system 30 that has been previously delayed. In detail, the diagnosis system 30 may include a failure detection unit 301 , a failure diagnosis unit 302 , and a failure prediction unit 303 , among which the failure detection unit 301 and the failure diagnosis unit 302 . ) may be referred to as FDD (Fault Detection & Diagnosis) or FDI (Fault Detection & Isolation).

The failure detection unit 301 receives information from the control system 10 or the power generation system 20 to determine which failure has occurred or is occurring in the plant, and the failure information is analyzed by the failure analysis unit 302 . It serves to transmit

The failure analysis unit 302 serves to perform various analyzes on failures of the plant based on the failure information received in advance.

Finally, the failure prediction unit 303 can be understood as a configuration that predicts a plant failure that may occur in the future based on the previously detected and analyzed contents, and provides the generated prediction information to the coordination system 40, The failure prediction unit 300 is also a subject to be mainly discussed in this detailed description. The types of objects that the failure prediction unit 303 can predict may include a plant signal, a plant failure time, a plant failure location, a plant alarm level, or a plant failure cause.

4 illustrates detailed configurations of the failure predicting unit 303 , the failure predicting unit 303 may further include a data processing unit 3031 , a predictive modeling unit 3033 , and a predicting unit 3035 .

The data processing unit 3031 is a configuration in which the predictive modeling unit 3033, which is a subsequent configuration, processes data suitable for learning, and more precisely, it can be understood as a configuration that filters only data suitable for learning in advance. In general, the amount of failure-related information collected from a plant is very large, and among these data, error values that are not correlated or that do not fit the situation are mixed, which can be a factor that reduces accuracy in subsequent predictive modeling. In this case, the data processing unit 3031 filters such unnecessary data in advance, thereby performing a role of increasing accuracy in future predictive modeling.

A specific data processing process of the data processing unit 3031 will be described later in more detail in the description of FIG. 5 , and here, the remaining components of the predictive modeling unit 3033 and the prediction unit 3035 will be further discussed.

The predictive modeling unit 3033 is a configuration that generates a failure prediction model that can be used for prediction operation in the prediction unit 3035, which is a subsequent component, and the predictive modeling unit 3033 may generate a failure prediction model according to various algorithms. However, as an example, a failure prediction model can be generated by using an artificial neural network. Artificial neural network is a data processing methodology that simulates inductive learning ability by mathematically modeling the information processing structure of the brain composed of nerve cells. Its basic purpose is to predict the output value of An artificial neural network consists of a parallel connection structure (layer) of nodes that play the role of nerve cells, and in general, serial connection of input layer-hidden layer-output layer is used. It is configured to have a plurality of hidden layers and may be implemented to handle complex input/output correlation. In the case of using such an artificial neural network, even in a situation where physical characteristics or relationships are not clearly known, correlation learning is possible only with inputs and outputs. It has the advantage of deriving a correlation.

Meanwhile, the predictive modeling unit 3033 may receive learning data from the data processing unit 3031 . At this time, the learning data may be in the form of a plurality of data sets, and each data set is It may include a normal data set having data characteristics, and a failure data set having data characteristics when the plant is in failure. A plurality of data sets will be described later.

As described above, the predictive modeling unit 3033 generates a model for predicting plant failure by using an arbitrary algorithm such as an artificial neural network, and the failure predictive model thus generated is provided to the predictive unit 3035, which is a subsequent configuration. It can be used to predict actual failure.

Next, the prediction unit 3035 receives information on the failure prediction model previously generated by the predictive modeling unit 3033, and applies the values for the current plant operating state to the failure prediction model to perform a simulation. By looking at it, the failure prediction values of the plant are calculated. At this time, the aforementioned 'values for the current plant operating state' are information received by the diagnostic system 30 from the control system 10 or information received by the diagnostic system 30 from the power generation system 20 . It may be my inclusion.

Meanwhile, the prediction unit 3035 may finally generate failure prediction information and transmit the failure prediction information to the coordination system 40 . The failure prediction information may include predicted values or predicted trends of values for various parameters, predicted failure time points, predicted failure locations, predicted alarm levels, predicted failure causes, and the like.

On the other hand, the failure prediction information generated by the prediction unit 3035 is not only transmitted to the adjustment system 40, but also provided to the data processing unit 3031 of the previous stage, that is, the plant failure prediction process is performed in the feedback loop. By having a shape, it is possible to enable more sophisticated failure prediction by additional filtering.

5 is for explaining each execution step of the data processing unit 3031 which will be mainly discussed in this detailed description.

The data processing unit 3031 first receives all information including at least one of the failure detection information, the failure diagnosis information, and the plant operation information ( S501 ). Failure detection information may include values related to failures detected and occurring in the plant, such as predicted values, residual values, or early warning values. As such, information that can be inferred by a series of calculations from the failure detection information may be included, and the plant operation information includes information indicating the current operation status such as plant output and efficiency, and whether the plant is currently in operation or the plant is in operation. Various measurements from the past may be included. Meanwhile, among the general information, failure prediction information may also be included, as briefly mentioned above.

After step S501, the data processing unit 3031 may perform a step S502 of filtering the general information according to a predefined filter condition. This filtering step can be performed according to the filter conditions defined based on the operator or user's basic knowledge. can be raised

After step S502, the data processing unit 3031 analyzes the operating state of the plant with reference to the overall information, more precisely, with reference to the failure history, and classifies it into a plurality of normal sections and failure sections. Step S503 is performed. In this case, the plurality of sections may be classified on the operating time of the plant, and more specifically, may be classified into a normal section in which the plant normally operates and a failure section in which a failure occurs in the plant in the operating time of the plant. A graph in which the horizontal axis is defined as time and the vertical axis as measured values is shown on the right side of step S503 in FIG. 5, which separately displays a section (failure section) when an arbitrary measured value of the plant has an abnormality in the operating time of the plant did it That is, in step S503, a failure section and a normal section other than the failure section on the time period displayed on the graph may be classified. On the other hand, it will be said that a standard is necessary to classify the normal section and the faulty section, and the standard at this time is preset by the operator or user (designer), or a diagnostic system that manages the execution of the data processing unit 3031 It may also be set by machine learning executable by the central processing unit of (30).

Meanwhile, after step S503, the data processing unit 3031 may analyze the plurality of sections classified in step S503 to generate a prediction data set (S504). The meaning of analyzing a plurality of sections may mean extracting the characteristics of data for each classified section. Multiple correlations such as whether the correlation (the high rate of failure of other devices when the first device failed) was high, or which part in the first device caused the most failures when the first device failed It can mean that the relationship is computed and extracted. As a result of this operation, a plurality of prediction data sets may be generated, and each data set may include a normal data set and a failed data set together as shown on the right side of step S504 of FIG. 5 . In this case, each data set may correspond to each device constituting the plant.

A method for predicting the failure of an abnormal plant and a system for this have been reviewed. The present invention is not limited to the specific embodiments and applications described above, and various modifications can be made by those of ordinary skill in the art to which the invention pertains without departing from the gist of the invention as claimed in the claims. Needless to say, these modifications should not be understood separately from the spirit or vision of the present invention.

10 control system
20 dynamometer
30 diagnostics
301 Failure detection unit 302 Failure diagnosis unit 303 Failure prediction unit
3031 data processing unit
3033 Predictive Modeling Department
3035 Prediction Unit
40 adjuster

Claims (15)

A method for predicting plant failure by a system having a central processing unit and a memory, the method comprising:
(a) receiving general information including at least one of failure detection information, failure diagnosis information, and plant operation information;
(b) classifying the operating state of the plant into a plurality of sections with reference to the general information; and
(c) generating a data set for prediction by analyzing the plurality of sections classified in step (b);
including,
The prediction data set is
including a normal data set and a faulty data set;
After step (c),
providing the generated prediction data set as learning data for failure prediction modeling;
further comprising,
A method of predicting failures in a plant.
According to claim 1,
After step (a),
filtering the general information according to a predefined filter condition;
further comprising,
A method of predicting failures in a plant.
3. The method of claim 2,
characterized in that the plurality of sections are classified on the operating time of the plant,
A method of predicting failures in a plant.
4. The method of claim 3,
The plurality of sections is characterized in that it is divided into a normal section and a failure section according to the operating state of the plant,
A method of predicting failures in a plant.
delete delete 5. The method of claim 4,
All information received in step (a) is,
It characterized in that it further comprises failure prediction information generated by the failure prediction modeling,
A method of predicting failures in a plant.
In the system for predicting plant failure as having a central processing unit and a memory,
The system includes a failure prediction unit,
The failure prediction unit,
Receives general information including at least one of failure detection information, failure diagnosis information, or plant operation information, classifies the operating state of the plant into a plurality of sections with reference to the general information, and divides the classified plurality of sections Analyze to create a data set for prediction,
The prediction data set includes a normal data set and a failure data set,
The generated prediction data set is provided to the failure prediction unit as learning data for failure prediction modeling,
Plant Failure Prediction System
9. The method of claim 8,
The failure prediction unit,
Characterized in filtering the general information according to a predefined filter condition,
Plant Failure Prediction System.
10. The method of claim 9,
characterized in that the plurality of sections are classified on the operating time of the plant,
Plant Failure Prediction System.
11. The method of claim 10,
The plurality of sections is characterized in that it is divided into a normal section and a failure section according to the operating state of the plant,
Plant Failure Prediction System.
delete delete 12. The method of claim 11,
The received general information is,
It characterized in that it further comprises failure prediction information generated by the failure prediction modeling,
Plant Failure Prediction System.
15. The method of claim 14,
A failure detection unit for detecting a failure of the plant; and
a failure diagnosis unit for analyzing the failure detected by the failure detection unit;
further comprising,
The failure prediction unit is characterized in that it receives at least a portion of the general information from any one of the failure detection unit or the failure diagnosis unit,
Plant Failure Prediction System.

Images