KR102406994B1 - Bigdata based system and method for diagnose and analyze abnormalities in facility - Google Patents

Abstract

In a big data-based facility abnormality diagnosis and analysis system performed by artificial intelligence that has learned a learning model composed of characteristic information based on time-series boiler status and fuel properties information, output and fuel change points within a predetermined range are normalized. Classify into states, establish a standard model of characteristic information according to output changes in a normal state, establish an evaluation model in which characteristic information reflecting the reference model is classified as abnormal, and A big data-based facility abnormality diagnosis and analysis system is provided that analyzes characteristic information between normal and abnormal states, analyzes a plurality of data change factors based on an evaluation model, and analyzes the sensitivity of characteristic information in an abnormal state. .

Description

Big data-based facility abnormality diagnosis and analysis system and method

The present invention relates to a system and method for diagnosing and analyzing equipment abnormalities based on big data.

In general, a system for monitoring the performance of a power plant refers to a performance monitoring system for a steam power plant or a combined cycle power plant for monitoring the performance of a steam power plant or a combined cycle power plant. Here, the system can be different depending on fuel information. In the case of energy, it means bullets, and in the case of complex, it can mean gas. It aims to derive the performance calculation results in real-time concept by building an interface for the connection between the database and the performance calculation engine reflecting this information and applying the data to the performance calculation engine. However, this conventional technique has a problem in that it is cumbersome to update the information in the database with the latest information, and it can be difficult to distinguish whether the current performance is in a normal state or an abnormal state only by numerically showing the current performance of the power plant. . Accordingly, there is a need for a system that not only manages the database, but also provides information that can more specifically and accurately classify the current state of the power plant.

Republic of Korea Patent Publication No. 10-2155344 (2020. 09. 07)

An embodiment of the present invention aims to construct a system and method that can be used without a failure mechanism and has a wider scope of application.

In a big data-based facility abnormality diagnosis and analysis system performed by artificial intelligence that has learned a learning model composed of characteristic information based on time-series boiler status and fuel properties information, output and fuel change points within a predetermined range are normalized. Classify into states, establish a standard model of characteristic information according to output changes in a normal state, establish an evaluation model in which characteristic information reflecting the reference model is classified as abnormal, and A big data-based facility abnormality diagnosis and analysis system is provided that analyzes characteristic information between normal and abnormal states, analyzes a plurality of data change factors based on an evaluation model, and analyzes the sensitivity of characteristic information in an abnormal state. .

In addition, the reference model may implement nonlinear characteristics formed by data through the nonlinear model.

In addition, whether or not the overheat reduction of the final reheater is involved, the characteristic information can be analyzed through pattern clustering of the characteristic information based on binary encoding.

In addition, if the overheating reduction factor of the final reheater intervenes, it is classified as an abnormal state, and the intervention of the overheating reduction unit may be made by a variable including an output out of the range of the normal state.

In addition, if it is less than 0.05, which is the statistical significance level (p-value) of the Abnormal Score, which is an index derived through the relationship between power output and fuel change time, it is classified as an abnormal state, and the abnormality index can be established by the following equation. can

,

째 특징정보의 정상상태값,

째 정상상태 표준편차)(here,

,

The steady state value of the second characteristic information,

second steady-state standard deviation)

final reheater overheat reduction final reheater overheat reduction final reheater overheat reduction

According to one embodiment of the present invention, it is possible to provide a system and method for diagnosing and analyzing equipment abnormalities based on big data that analyzes the overall health of equipment by using big data and performs cause diagnosis based on this.

According to an embodiment of the present invention, big data that can improve the utilization of dark data of power plants that have not been used in the past and compensate for the disadvantages of the existing approach through data-based comprehensive health monitoring and diagnostic analysis technology of facilities/systems It is possible to provide a system and method for diagnosing and analyzing equipment abnormalities based on the present invention.

1 is a flowchart showing a driving sequence (ie, a driving method) of a big data-based facility abnormality diagnosis and analysis system according to an embodiment of the present invention;
2 is a diagram showing an example of an area for extracting feature information according to an embodiment of the present invention in a facility;
3 is a graph in which feature information is displayed as a normalization model (reference model) through a plurality of steady states according to an embodiment of the present invention;
4 is a graph showing an evaluation model implemented by collecting characteristic information of an abnormal state based on a normal state according to an embodiment of the present invention;
5 is a view showing binarization to classify normal and abnormal in each area during operation of a facility according to an embodiment of the present invention;
6 is a view showing that factors affecting the abnormal state are collected as a plurality of data according to an embodiment of the present invention;
7 is a graph illustrating measuring sensitivity by adjusting the plurality of data according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. However, this is merely an example, and the present invention is not limited thereto.

In the description of the present invention, if it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. And, the terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to intentions or customs of users and operators. Therefore, the definition should be made based on the content throughout this specification.

The technical spirit of the present invention is determined by the claims, and the following examples are only one means for efficiently explaining the technical spirit of the present invention to those of ordinary skill in the art to which the present invention belongs.

Hereinafter, as an embodiment of the present invention, a system capable of detecting an inefficient operation or an abnormal state in a power generation facility will be described. The above embodiment is a big data-based facility abnormality diagnosis and analysis system (hereinafter referred to as the system) and its operating method performed by artificial intelligence learning a learning model composed of characteristic information based on time-series boiler status and fuel property information More effective operation of power generation facilities can be promoted by detecting abnormal conditions and inefficient operation by artificial intelligence that has learned through the collected data and to improve them. Here, the steady state may be distinguished from the abnormal state based on the output of the predetermined range and the time of fuel change.

In addition, it can be divided into stages such as normal, caution, and danger based on the opening degree and holding time of the overheat reduction valve of the final reheater, and data related to main variables indicating the boiler operation status and heat balance of the water supply or steam system are extracted And by processing, calculating an average value in a predetermined time unit, and extracting features from the result value, it is possible to produce final data to be provided to the user.

1 is a flowchart illustrating a driving sequence (ie, a driving method) of a big data-based facility abnormality diagnosis and analysis system according to an embodiment of the present invention, and FIG. 2 is a flowchart illustrating extracting feature information according to an embodiment of the present invention. It is a diagram showing an example of the area to be used in the facility.

Referring to FIGS. 1 and 2 , the time of changing the output and fuel may be divided into a normal state and an abnormal state through the range determined in advance. Subsequently, the reference model establishment step, the evaluation model establishment step, the evaluation model analysis step, and the sensitivity analysis step may be sequentially performed. In addition, the characteristic information analysis step may be performed in parallel with the sensitivity analysis step and the characteristic information analysis step after the evaluation model establishment step. Information obtained from the preceding steps may be collected and provided to the user.

Specifically, a standard model of characteristic information according to output change in a steady state is established, an evaluation model in which the characteristic information reflecting the reference model is classified into an abnormal state is established, and the normal state based on the intervention of the and a big data-based facility abnormality diagnosis and analysis system that analyzes characteristic information between abnormal states, analyzes a plurality of data change factors based on an evaluation model, and analyzes the sensitivity of characteristic information in an abnormal state is provided.

Here, the reference model may implement a nonlinear characteristic formed by data through a nonlinear model. For example, it can be implemented through a local polynomial regression model.

In addition, whether or not the overheat reduction of the final reheater is involved, the characteristic information can be analyzed through pattern clustering of the characteristic information based on binary encoding. That is, whether the final reheater overheating reduction agent is involved may be one of various characteristic information that may occur in the operating conditions during operation of the power generation facility. When the overheating reduction factor of the final reheater intervenes, it is classified as an abnormal state, and the intervention of the overheating reduction unit can be determined as being made by various variables including the output out of the range of the normal state.

In addition, if it is less than 0.05, which is the statistical significance level (p-value) of the Abnormal Score, which is an index derived through the relationship between power output and the time of fuel change, it is classified as an abnormal state, and the abnormality index can be established by the following equation. can

,

째 특징정보의 정상상태값,

째 정상상태 표준편차일 수 있다. here,

,

The steady state value of the second characteristic information,

The second may be the steady-state standard deviation.

The above-described driving and driving conditions may be applied to a water supply or steam system, such as the first area A1, the second area A2, and the third area A3 shown in FIG. 2 . In a more expanded sense, it may include the supply of fuel, combustion of the fuel, and a configuration in which an output is generated by burning the fuel. This includes various variables such as output such as output and input such as fuel supply. This is in order to be able to control or prepare for it.

3 is a graph in which feature information is displayed as a normalization model (reference model, FRM; Feature Reference Model) through a plurality of steady states according to an embodiment of the present invention.

Referring to FIG. 3 , the operating conditions selected within the range of a steady state are schematically shown. The horizontal axis of each graph means normalized output, and the vertical axis means normalized characteristic information value. On the graph, the area indicated by gray dots is determined to be operation in a normal state, and the area indicated by red dots is determined to be operation in an abnormal state. Also, the blue solid line may indicate a reference model.

Through this, it can be seen that the normal state and the abnormal state are clearly distinguished. At least, compared to the gray points distributed around the reference model, it can be seen that the distribution area of the red dot greatly deviates from the blue solid line indicating the reference model.

4 is a graph illustrating an evaluation model implemented by collecting characteristic information of an abnormal state based on a normal state according to an embodiment of the present invention.

Referring to FIG. 4 , it may be an evaluation model (ASM; Abnormal Score Model) constructed through characteristic information revealed during driving in an abnormal state. This may be a model formed by collecting characteristic information generated during operation in an abnormal state by using the reference model described above.

That is, it may be a model formed by constructing a reference model and collecting characteristics generated when an operation determined to be an abnormal operation by comparison with the reference model is performed. In other words, various variables including operating conditions and outputs revealed from these evaluation models can act as factors that induce abnormal operating conditions. Accordingly, the following FIG. 6 shows that factors that induce such driving conditions are specified and collected as characteristic information.

Meanwhile, when the evaluation model is formed, the normal state or the abnormal state in the first area A1 , the second area A2 , and the third area A3 can be separately displayed with reference to FIG. 2 . This can be displayed through binary encoding as shown in FIG. 5 . 5 is a binarized view for classifying a normal or abnormal state in each area during operation of a facility according to an embodiment of the present invention. This may be divided based on the power of the predetermined range of the present invention described above and the fuel change time, which may be a criterion selectively determined by the user.

6 is a diagram illustrating a collection of factors affecting an abnormal state as a plurality of data according to an embodiment of the present invention.

Referring to FIG. 6 , when the evaluation model is analyzed, factors causing the characteristic information can be analyzed through characteristic information of an abnormal state revealed in the evaluation model as described above. In this example, 29 factors were specified, and the factors may be factors that cause some differences in statistical significance (P-value) between the steady state and the abnormal state. These may be factors selected in consideration of the design characteristics and the relationship between the system and the system that affect the operation of the final reheater overheating reduction (intervention).

Therefore, factors that show a difference of more than a predetermined numerical value quantitatively between the steady state and the abnormal state can be collected, but factors that are determined to have an influence of more than a certain level in indirectly inducing the abnormal state are also collected. That is, factors determined to be significant in terms of physical and logical flow of data may be included in the feature information factor. 7 , it is shown that each factor is controlled while measuring how much influence the factor has on the driving state and the sensitivity can be recognized.

7 is a graph illustrating measuring sensitivity by adjusting the plurality of data according to an embodiment of the present invention.

Referring to FIG. 7 , for example, in a low output state, 29 factors included in the characteristic information were measured by exposing the power generation facility to a low output (280 to 390M/W) state and a high output (390 to 500M/W) state. .

number Characteristic information variable name Sensitivity Influence low power
(280~390M/W) high power
(390 ~ 500M/W) One MW - - - 2 PE_IH_AMT M M L 3 SE_O_AST H H L 4 ST_O1_ST M L M 5 ST_O3_ST L L L 6 ST_O5_ST M M L 7 SP_O_AST M M L 8 DSH_IL1_ST M M L 9 DSH_O1_ST L L L 10 PSH_IA_AST H H L 11 PSH_OA_AT M M L 12 PSH_IB_AST H H L 13 PSH_OB_AT H H L 14 FSH_IA_ST H H L 15 FSH_OHA_AST H L H 16 FSH_IB_AST H H L 17 FSH_OHB_AST M M L 18 HR_IA_AT L L L 19 HR_OA_AT L H H 20 HR_IB_AT L L L 21 HR_OB_AT L H H 22 D_EC H H L 23 D_WW H H L 24 D_DSH H H L 25 D_PSHA H H L 26 D_PSHB H H L 27 D_FSHA M H M 28 D_FSHB H H L 29 D_HRA L H H 30 D_HRB L H H

As an example as shown in Table 1 above, when soundness deviation occurs under low power conditions, variables with high influence are PSH_IA_AST, PSH_IB_AST, PSH_OB_AT, FSH_IA_ST, FSH_IB_AST, D_EC, D_WW, D_DSH, D_PSHA, D_PSHB, D_FSHA, and D_FSHB. , when soundness deviation occurs under high output conditions, the variable with high influence except for the final reheater was identified as FSH_OHA_AST.

As with these experimental values, when an abnormal driving condition proceeds in a specific driving environment, it can be identified as a result of a specific factor among the characteristic information and can be dealt with.

Although representative embodiments of the present invention have been described in detail above, those of ordinary skill in the art to which the present invention pertains will understand that various modifications are possible within the limits without departing from the scope of the present invention with respect to the above-described embodiments. . Therefore, the scope of the present invention should not be limited to the described embodiments, and should be defined by the claims described below as well as the claims and equivalents.

S10: Reference model collection step
S20: Evaluation model establishment stage
S30: Evaluation model analysis step
S40: Sensitivity analysis step
S50: feature information analysis step
A1: first area
A2: Area 2
A3 : 3rd area

Claims (10)

In a big data-based facility abnormality diagnosis and analysis system performed with artificial intelligence that has learned a learning model composed of characteristic information based on time-series boiler status and fuel property information,
The output and fuel change time within a predetermined range are divided into normal conditions,
Establishing a reference model of characteristic information according to output change in the steady state,
Establishing an evaluation model in which the characteristic information reflecting the reference model is classified into an abnormal state,
After the establishment of the evaluation model, the characteristic information between the normal state and the abnormal state is analyzed based on the intervention of the final reheater overheating reduction factor,
Analyze a plurality of data change factors based on the evaluation model,
Analyze the sensitivity of the feature information in the abnormal state,
If it is less than 0.05, which is a statistical significance level (p-value) of an Abnormal Score, which is an index derived through the relationship between the output and the fuel change time, it is classified as the abnormal state, and the abnormal index is

A big data-based facility abnormality diagnosis and analysis system that establishes consciousness.
(here,

,

The steady state value of the second characteristic information,

second steady-state standard deviation)
The method according to claim 1,
The reference model is
A big data-based facility abnormality diagnosis and analysis system that implements the non-linear characteristics formed by data through a non-linear model.
The method according to claim 1,
Whether the final reheater overheating reduction is intervened,
A big data-based facility abnormality diagnosis and analysis system that analyzes the characteristic information through pattern clustering of the characteristic information based on binary encoding.
The method according to claim 1,
When the final reheater overheat reduction water intervenes, it is classified into the abnormal state, and the overheat reduction unit is intervened by a variable including an output out of the range of the steady state, a big data-based facility abnormality diagnosis and analysis system.
delete In a big data-based facility abnormality diagnosis and analysis system performed with artificial intelligence that has learned a learning model composed of characteristic information based on time-series boiler status and fuel property information,
The output and fuel change time within a predetermined range are divided into normal conditions,
Establishing a reference model of characteristic information according to output change in the steady state,
Establishing an evaluation model in which the characteristic information reflecting the reference model is classified into an abnormal state,
After the establishment of the evaluation model, the characteristic information between the normal state and the abnormal state is analyzed based on the intervention of the final reheater overheating reduction factor,
Analyze a plurality of data change factors based on the evaluation model,
Analyze the sensitivity of the feature information in the abnormal state,
If it is less than 0.05, which is a statistical significance level (p-value) of an Abnormal Score, which is an index derived through the relationship between the output and the fuel change time, it is classified as the abnormal state, and the abnormal index is

A big data-based facility abnormality diagnosis and analysis method that establishes consciousness.
(here,

,

The steady state value of the second characteristic information,

second steady-state standard deviation)
7. The method of claim 6,
The reference model is
A big data-based facility abnormality diagnosis and analysis method that implements the non-linear characteristics formed by data through a non-linear model.
7. The method of claim 6,
Whether the final reheater overheating reduction is intervened,
A big data-based facility abnormality diagnosis and analysis method for analyzing the characteristic information through pattern clustering of the characteristic information based on binary encoding.
7. The method of claim 6,
If the final reheater overheat reduction water intervenes, it is classified into the abnormal state, and the overheat reduction unit is intervened by a variable including an output out of the range of the steady state, a big data-based facility abnormality diagnosis and analysis method.
delete

Images