KR20220059193A - System and Method for providing early warning to prevent false alarm - Google Patents

Abstract

Disclosed is an early warning system capable of detecting mysterious symptom of a plant in advance. The early warning system includes: a collection unit which collects data of power generation facilities; a calculation unit which calculates data for each signal using the data of the power generation facilities, performs dimension reduction and data visualization through a recommended learning interval using the data for each signal to set a final learning interval, and updates a target model to execute model retraining, which creates an updated model; and a prediction unit for generating prediction information for real-time prediction of the power generation facilities in order to prevent false detection by using the update model.

Description

BACKGROUND ART System and Method for providing early warning to prevent false alarm

The present invention relates to early warning technology, and more particularly, to an early warning system and method for detecting abnormal signs of a plant in advance.

In addition, the present invention relates to an early warning system and method capable of preventing frequent false alarms due to incorrect selection of a learning section when generating a predictive model that is the core of early warning performance.

In general, large-scale plants, such as power generation or chemical, are operated in which hundreds of mechanical and electrical facilities are intricately connected. In order to stably supply power and products to such large plants, reliability must be secured by constantly measuring abnormal signs that cause accidents.

The performance of the predictive model, which is the core of early warning technology, is to extract a signal group with high physical correlation based on past history data, learn the steady state pattern of the signal group, and then change the signal to an abnormal state outside the learned normal section. Early detection of trends.

However, in the case of the existing machine learning method, the user must directly select an arbitrary section by relying on a chart or graph for historical data each time. Moreover, even if a predictive model is generated, history management is not performed on whether there has been a similar signal pattern in the past or how much data is required for learning. Therefore, even during re-learning, the same operation must be repeated, and an appropriate learning section can be found through trial and error.

For this reason, in the prior art, when the learning section is incorrectly selected, the signal pattern of the current state is not learned, and thus a number of false alarms are generated. This results in aggravating the confusion of the system operator and increasing user's inconvenience and/or work due to an additional operation for checking whether an actual alarm is made.

In particular, for power generation facilities, maintenance is essential for long-term operation, and regular/irregular maintenance work occurs. In general, the condition of the facility changes after maintenance, and re-learning is required to reflect the changed condition.

In addition, since the equipment state can change depending on the season and load conditions, and an update of the machine learning model is required, an early warning system capable of strengthening the function of model correction and minimizing false alarms is required.

1. Korea Patent No. 10-1960754 (Registration Date: 2019.03.15)

The present invention has been proposed to solve the problems according to the above background art, and an object of the present invention is to provide an early warning system and method capable of detecting abnormal signs of a plant in advance.

Another object of the present invention is to provide an early warning system and method capable of preventing frequent false alarms in advance due to selection of an erroneous learning section when a prediction model, which is the core of early warning performance, is generated.

The present invention provides an early warning system capable of detecting abnormal signs of a plant in advance in order to achieve the above object.

The early warning system is

a collection unit for collecting data of power generation facilities;

Calculating data for each signal using the data of the power generation facility, setting the final learning section by performing dimension reduction and data visualization through the recommended learning section using the data for each signal, and updating the target model to update the model a calculation unit that executes model re-learning to generate ; and

and a prediction unit that generates prediction information for real-time prediction of the power generation facility in order to prevent false detection by using the update model.

In addition, the data for each signal is an average value of current operation data for each signal for the data of the power generation facility and an average value of past data for each signal of the selected past inquiry period.

In addition, based on the average value of the current driving data for each signal, the similarity with the average value of the historical data for each signal within the past inquiry period is calculated.

In addition, the similarity is characterized in that it is calculated using a distance scale including a Euclidean distance, a Maharanovis distance, and a Manhattan distance.

In addition, the recommended learning interval is calculated using a final recommendation period determined when the month coincides in two or more distance scales among the distance scales through voting for the month with the highest similarity according to the ranking of the similarity. do.

In addition, the dimensionality reduction is characterized in that the data for each signal is converted into a multivariate dimension of a three-dimensional space by performing a dimensional reduction in a low dimension.

In addition, the dimensional reduction is a technique of finding an axis that minimizes the mean square distance between a high-dimensional data set and a projected data set so that the variance of the data for each signal is maximized, or a high-dimensional data set by learning the distribution of the data for each signal It is characterized in that it is performed using a technique of projecting data according to a decision boundary that optimizes the separation between the two.

In addition, the high-dimensional data set includes an entire data set for a period arbitrarily inquired by a user, a recommended data set recommended from average value-based similarity measurement, and a current operation data set for the last changed state of the power generation facility. do it with

In addition, the size of the high-dimensional data set is determined using kurtosis, which is a measure indicating the sharpness of the probability distribution from the center calculated by finding the center of the recommended data set and calculating the range of data with high cluster density from the center. It is characterized in that the distribution is extracted as a data range of a form close to a normal distribution.

In addition, the size of the high-dimensional data set is characterized in that it is selected through a user's drag-and-drop method.

In addition, in order to prevent a false alarm that may be caused by a signal pattern that may not be learned in the learning section, it is characterized in that a variable threshold value whose value is changed is set.

In addition, in order to prevent a false alarm that may be caused by a signal pattern that may not be learned in the learning section, a static threshold value at which a value is fixed is set.

On the other hand, another embodiment of the present invention, comprising the steps of: (a) collecting the data of the power generation equipment; (b) the calculation unit calculates data for each signal using the data of the power generation facility, performs dimension reduction and data visualization through a recommended learning section recommended using the data for each signal, sets the final learning section, and sets the target model executing model retraining to update the update model; and (c) generating, by a prediction unit, prediction information for real-time prediction of the power generation facility to prevent false detection by using the update model. .

According to the present invention, not a complicated algorithm, but statistical values and the degree of data clustering are measured by the degree of similarity (Euclidean, Maharanovis, Manhattan distance), and an optimal data set is obtained using the voting result, and it is predicted By using it as the training data of the model, the performance of the predictive model can be improved and false alarms can be prevented in advance.

In addition, as another effect of the present invention, by reducing the dimension of the variable, the history section of the machine learning data set that affects the machine learning performance of early warning is visually confirmed, and the change of season, change of equipment condition due to maintenance, and external environment For example, it is easy to determine whether the model needs re-learning according to changes in conditions (changes in ambient temperature, etc.).

In addition, as another effect of the present invention, it is possible to secure sophisticated and optimized learning data for each facility when establishing a long-term DB by securing patterns of learning data.

In addition, as another effect of the present invention, it is possible to selectively introduce and reuse a dataset of an existing model in which the current driving data has a similar pattern in the past, thereby minimizing the frequently occurring model modification work and increasing the efficiency of model management. The point is that learning performance and learning speed can be improved by optimizing data size by removing unnecessary learning period.

1 is a block diagram of an early warning system for preventing false detection according to an embodiment of the present invention.
2 is a flowchart illustrating an early warning implementation process for preventing false detection according to an embodiment of the present invention.
3 is a table showing a trigger operation using an event frequency according to an embodiment of the present invention.
4 is a conceptual diagram illustrating a dimension reduction for data change visualization using two dimensions according to an embodiment of the present invention.
5 is a visualization graph using dimension reduction according to an embodiment of the present invention.
6 is a visualization graph of a learning section proposal according to an embodiment of the present invention.
7 is a graph showing tracking management of learning data according to an embodiment of the present invention.
8 is a view showing facility history data according to an embodiment of the present invention.
9 is a graph showing a result of a recommended data set for a facility according to an embodiment of the present invention.
10 is a table showing a recommendation result of similarity-based learning data according to an embodiment of the present invention.
11 to 13 are graphs showing visualization of a power generation facility according to an embodiment of the present invention.
14 is a diagram illustrating size selection using a distribution of a training data set according to an embodiment of the present invention.
15 is a diagram illustrating size selection using a distance from the center of a training data set according to an embodiment of the present invention.
16 is a diagram illustrating information provision for model correction according to an embodiment of the present invention.
17 is a diagram showing tracking of a learning data pattern according to an embodiment of the present invention.
18 is a diagram illustrating occurrence of false positives due to a static threshold according to an embodiment of the present invention.
19 is a diagram illustrating occurrence of a false detection due to a variable threshold according to an embodiment of the present invention.
20 is a graph comparing before and after optimization according to an embodiment of the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and will be described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

In describing each figure, like reference numerals are used for like elements. Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. The term “and/or” includes a combination of a plurality of related listed items or any of a plurality of related listed items.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

Terms such as those defined in commonly used dictionaries should be interpreted as having meanings consistent with the meanings in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. shouldn't

Hereinafter, an early warning system and method for preventing false detection according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an early warning system 100 for preventing false detection according to an embodiment of the present invention. Referring to FIG. 1 , the early warning system 100 includes a communication unit 110 that receives data about power generation facilities, a collection unit 120 that collects and stores data of power generation facilities, and uses the data to collect data for each signal. Calculation unit 130 that calculates and updates the target model to update the target model by performing dimensionality reduction and data visualization through the recommended learning section using data for calculation and data visualization, false detection using the update model It may be configured to include a prediction unit 140 for generating prediction information for predicting in real time, an output unit 150 for outputting prediction information, and the like.

The communication unit 110 performs a function of being connected to another external communication device (not shown) through a communication network (not shown). Communication can be wired or wireless. In addition, the communication unit 110 may use various communication protocols for wireless communication and the like. To this end, the communication unit 110 may include a modem, a microprocessor, a communication circuit element, and the like.

A communication network refers to a connection structure that enables information exchange between each node, such as a plurality of terminals and servers, and includes a public switched telephone network (PSTN), a public switched data network (PSDN), and an integrated services digital network (ISDN) Networks), Broadband ISDN (BISDN), Local Area Network (LAN), Metropolitan Area Network (MAN), Wide LAN (WLAN), etc., but , the present invention is not limited thereto, and the present invention is not limited thereto, and wireless communication networks such as CDMA (Code Division Multiple Access), WCDMA (Wideband Code Division Multiple Access), Wibro (Wireless Broadband), WiFi (Wireless Fidelity), HSDPA (High Speed Downlink Packet Access) It may be a network, Bluetooth, Near Field Communication (NFC) network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, and the like. Alternatively, it may be a combination of these wired communication networks and wireless communication networks.

Examples of the external communication device include a server, a personal computer (PC), a notebook computer, and a notepad. Of course, it is also possible to transfer data stored in an external communication device to the USB without passing through such a communication network.

The collection unit 120 performs a function of collecting data of the power generation facility transmitted through the communication unit 110 . The data of the power generation facility may be transmitted from the power generation facility, or may be transmitted through an external communication device that is primarily generated and stored secondarily in the power generation facility. The collection unit 120 stores the transmitted data in the database DB of the storage unit 160 .

The calculation unit 130 calculates data for each signal by using the data, performs dimension reduction and data visualization through a recommended learning section using the data for each signal, determines the learning section, updates the target model, and generates an update model perform the function In addition, the calculator 130 recommends an optimal learning section through similarity measurement, and performs a function of verifying current driving data and recommended learning data through dimension reduction and visualization. In addition, a database of optimal learning patterns is secured through tracking and management of the learning data, and the size of the learning data is optimized using a distance from the center or a statistical value. In addition, it implements false detection prevention by applying a variable threshold.

The prediction unit 140 performs a function of generating prediction information for predicting a false positive in real time by using the update model. In addition, it monitors the distance between the learning data center and the current driving data center, and generates an alarm when a distance increase occurs.

The output unit 150 performs a function of displaying information processed by the collection unit 120 , the calculation unit 130 , the prediction unit 140 , and the like. Accordingly, the output unit 150 may output information in a combination of text, voice, and graphics. To this end, the output unit 150 may be configured to include a display, a sound system, and the like.

The display may be a liquid crystal display (LCD), a light emitting diode (LED) display, a plasma display panel (PDP), an organic LED (OLED) display, a touch screen, a cathode ray tube (CRT), a flexible display, or the like. In the case of a touch screen, it may function as an input means.

The storage unit 160 performs a function for storing software, an instruction set, or data for operating the system 100 . To this end, the storage unit 160 is a random access memory (Random Access Memory, RAM), a magnetic disk (Magnetic Disc), a flash memory (Flash Memory), a static RAM (Static Random Access Memory, SRAM), a ROM (Read Only Memory, ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), etc., but is not limited thereto.

The collection unit 120 , the calculation unit 130 , and the prediction unit 140 illustrated in FIG. 1 mean a unit that processes at least one function or operation, which may be implemented as software and/or hardware. In hardware implementation, application specific integrated circuit (ASIC), digital signal processing (DSP), programmable logic device (PLD), field programmable gate array (FPGA), processor, microprocessor, other It may be implemented as an electronic unit or a combination thereof.

In software implementation, software composition component (element), object-oriented software composition component, class composition component and task composition component, process, function, attribute, procedure, subroutine, segment of program code, driver, firmware, microcode, data , databases, data structures, tables, arrays, and variables. Software, data, etc. may be stored in a memory and executed by a processor. The memory or processor may employ various means well known to those skilled in the art.

2 is a flowchart illustrating an early warning implementation process for preventing false detection according to an embodiment of the present invention. Referring to FIG. 2 , when a false alarm frequently occurs within a certain time in a plurality of predictive models due to a change in the state of a power generation facility due to maintenance or seasonal change, a trigger function is used to provide the model to the user. Re-learning is requested (step S210). An example showing a trigger operation through an event frequency for requesting such model retraining is shown in FIG. 3 . This will be described later.

If there is a request for model re-learning, a procedure ( S200 ) of processing data to execute model re-learning is executed. The model re-learning procedure ( S200 ) may consist of steps S220 to S270 .

First, the average value of current operation data for each signal is calculated for the data of the power generation facility acquired through the collecting unit 120 (steps S220 and S230). In the case of calculating the average value of the current driving data for each signal, the user inquires the entire period required for learning, and the development system calculates the average value for each signal (tag) for the current driving data and the monthly average value for the selected past inquiry period. . This is shown in a table as follows.

In the table above, columns "P1LAB~" and the like represent signals (tags), and rows represent monthly average calculations for the inquiry period for each signal. In addition, the present value represents an average value of the current driving data.

The above table is a tag for BFP (boiler feed pump, boiler feed water pump), and BFP corresponds to the feed water system of the power generation facility and an auxiliary facility that supplies water to the boiler to generate steam. The BFP consists of a pump and a motor, and the signals indicated above are the inlet/outlet temperature of the BFP, the discharge pressure of the inlet/outlet, and the bearing vibration and temperature signals of the pump and the motor, respectively. It indicates the operation data of the facility.

In other words, when the user sets the inquiry period for past history data required for learning (about 1 to 3 years), the system calculates the average value for each signal (tag) included in the model from maintenance to the present time, For the inquiry section, the average value of data for each signal (tag) is calculated for each month.

Thereafter, the average value and similarity of the data for each monthly signal (tag) within the inquiry period (for example, about 3 years, 2016-10 ~ 2019-10) is measured based on the average value of the current driving data of each signal (tag) ( step S240). In this case, the measurement method may be a distance scale such as a Euclidean distance, a Maharanovis distance, or a Manhattan distance.

The equations for Euclidean distance, Mahalanobis Distance, and Manhattan distance are as follows.

Euclidean distance is a method of measuring the linear distance between two data (x,y) (vector).

The Mahalanobis distance is a method of expressing how many times the standard deviation is the distance from the mean. Therefore, when the covariance of the data is 0 as the covariance (σ) calculation is added to the Euclidean distance, the Mahalanobis and the Euclidean distance are the same. Here, T denotes a transformation matrix.

The Manhattan distance is a method that represents the sum of the absolute values of the shortest distances connecting two vectors (x,y).

Therefore, for the inquiry period, the distance value between the average value of the current driving data and the average value of the past data for each signal is calculated as Euclidean distance, Mahalanobis distance, and Manhattan distance, respectively. The table below summarizes this.

In the table above, the top row represents the month of the current driving data, and the leftmost column represents the month of the past data.

After that, the recommendation of the optimal learning section through voting is made (step S250). In other words, a similarity ranking is calculated through three distance scales, and the month with the highest similarity is voted according to the similarity ranking. That is, the entire history data section having the highest similarity is recommended using the total calculated distance value. If the recommendation period (months) matches in two or more techniques among the three distance scales, the learning period is delivered to the user as the final recommendation period.

In addition, even when the user wants to be recommended for data for two months or more, the monthly ranking is selected using the similarity and learning data is recommended according to the voting result.

After that, the user needs to visually confirm the final recommendation period. At this time, since it is difficult to check all the tags as a target, the data is converted into a three-dimensional multivariate dimension by reducing the dimension to a low dimension ( step S260). By visualizing the entire history data for the inquiry period (i.e., the past data set for the period arbitrarily inquired by the user), the recommended learning section (i.e., the data set recommended by the voting result), the status of the current driving data in a three-dimensional space expression, and after confirming this, the user confirms the recommended learning section setting (step S261).

Of course, if the recommended learning section is not determined in step S261, steps S230 to S260 proceed again.

Meanwhile, by setting a variable threshold, a function of systematically supplementing false alarms that may be caused by a signal pattern that may not be learned in the learning section may be performed (step S270) .

In step S261, when the recommended learning section is determined as a result of confirmation, an update is performed on the target model to generate an update model (step S280).

Thereafter, real-time prediction of the power generation facility is performed using the updated model (step S290).

On the other hand, when the update model is generated, the learning data center - the current driving data center distance is monitored (step S280). Monitoring is the current operation of the same signal (tag) with the center (average) of the reconstructed data set after dimension reduction for the data set selected for training among the past data for a plurality of signals (tags) participating in the model After dimensionality reduction of the data, the distance from the center (average) of the reconstructed data set is monitored.

Thereafter, as the distance between the center of the learning data and the center of the current driving data increases, it means that the current state is different from the learning pattern. If the center-to-center distance is greater than the value set by the user and the alarm frequently occurs in the early warning system, the user can be informed that the prediction model needs to be updated. That is, the user can prevent false alarms in advance by updating the model using the increase in the center distance and the alarm frequency of the early warning system (step S281).

3 is a table showing a trigger operation using an event frequency according to an embodiment of the present invention. Referring to FIG. 3 , when a season change or a facility state change after maintenance, an alarm of the same signal (Tag) is continuously generated, which is not an alarm due to an abnormal symptom of the actual facility, but rather the condition of the facility. It is a false detection that occurs because the learning pattern of the predictive model and the current pattern are different due to the change. Therefore, it can be seen that re-learning of the existing model is necessary. Therefore, the user is requested to relearn the model using a trigger based on the target model and the frequency of alarm occurrence.

4 is a conceptual diagram illustrating a dimension reduction for data change visualization using two dimensions according to an embodiment of the present invention. In general, since there are various facilities in a large plant such as a power plant, 20 or more machine learning models must be operated for each facility. Due to this, the number of signals (tags) to be managed becomes 20×20~30 = 400~600. Therefore, as the number of equipment increases, the number of management tags rapidly increases, so it is difficult to efficiently manage the model in the conventional method.

In addition, in order to check the signal change with respect to a plurality of signals (tags) in the model, it is inconvenient to check all signals one by one using a two-dimensional graph. Dimensional reduction is carried out to solve this inconvenience. Among the high-dimensional data characteristics, it is possible to express high-dimensional data with some characteristics, and it is a method of reducing the dimension by projecting it into a low-dimensional space (410, 420, 430, 440). Referring to FIG. 4 , for dimensionality reduction, a technique for finding an axis that minimizes the mean square distance between a high-dimensional data set and a projected data set so that the dispersion of data is maximized, or by learning the distribution of data between high-dimensional data sets Depending on the decision boundary that optimizes the separation, a technique for projecting the data can be introduced.

The high-dimensional data set to be expressed visually is divided into three types. First, the entire data set of the period arbitrarily inquired by the user, second, the recommended data set recommended from the average value-based similarity measurement, and finally the current operation data set for the changed state of the power generation facility should be visually plotted. As described above, there are too many tags (dimensions) in one predictive model, so it is difficult to schematize them in one graph. To this end, it is possible to check the degree of clustering for all tags with reduced dimensions in the three-dimensional space reduced to the low dimension as described above.

5 is a visualization graph using dimension reduction according to an embodiment of the present invention, and FIG. 6 is a visualization graph of a learning section proposal according to an embodiment of the present invention. 5 and 6 , as a method for minimizing learning data while maintaining high learning performance, the center of the recommended dataset is found, and the range of data with high cluster density from the center is calculated, and the sharpness of the probability distribution is calculated from the center. The size of the data set can be optimized by extracting a data range whose distribution is close to a normal distribution using kurtosis, which is a measure that represents it. The distance of the learning data may be arbitrarily adjusted according to the user's judgment, and the size of the learning data may be changed according to the selected distance.

7 is a graph showing tracking management of learning data according to an embodiment of the present invention. Referring to FIG. 7 , the current driving data set and the recommended data set can be checked monthly or selectively, and it is determined whether the data set recommended by the system is set as training data for model update. In addition, it is possible to provide information on the optimal learning pattern during long-term operation of the system by tracing and managing the movement path centered on the learning data.

8 is a view showing facility history data according to an embodiment of the present invention. In particular, FIG. 8 is historical data for a CEP (Changjo Energy Plant) facility among power plant auxiliary equipment (balance of plant) facilities. Referring to FIG. 8 , it can be seen that the maintenance of the power generation facility occurred three times during the inquiry period, and the signal pattern slightly changed according to the state of the power generation facility before and after the maintenance.

9 is a graph showing a result of a recommended data set recommended for a facility according to an embodiment of the present invention. Referring to FIG. 9 , in order to examine the effectiveness of the similarity method using distance, 10 months were randomly selected from the past data sets (July 2016 to April 2018) as the current driving data set for the CEP facility. For the 10 selected cases, it was reviewed whether the data set recommended by the three similarity techniques recommends the same month at least two of the data sets through voting. It was confirmed that two or more similarity methods recommended the same month for 10 cases.

10 is a table showing a recommendation result of similarity-based learning data according to an embodiment of the present invention. Referring to FIG. 10, as well as CEP (Condensate Extraction Pump) equipment, BFP (Boiler Feedwater Pump), CWP (Circulating Water Pump) pumps, FAB (Fluidizing Air Blower), IDF (Induced Draft Fan), PAF (Primary Air) Fan) and SAF (Secondary Air Fan) were selected for training data to prevent false alarms for the same inquiry period. In each facility, a total of 28 cases can be confirmed as the same facility is configured in duplicate and is installed in two units respectively.

The result of measuring the similarity (Euclidean distance, Mahalanobis distance, Manhattan distance) using the historical data set (monthly) and the average value after setting the current operation data set for each facility for the inquiry period from July 2016 to April 2018 , it can be seen that about 92% of all cases recommended the same period (month) as the training data, and only 8% recommended three different sets of data (months) with different degrees of similarity.

Some of these include cases in which the current state forms a completely different data cluster from the past historical data due to a change in the actual equipment state.

11 to 13 are graphs showing visualization of a power generation facility according to an embodiment of the present invention. In particular, FIG. 11 is a visualization of the IDF facility of Unit 1. The entire high-dimensional data set 1110 is visualized as a recommendation data set, a past data set, and a current driving data set 1120 .

12 is a visualization of the SAF facility at Unit 1. The entire high-dimensional data set 1210 is visualized as a recommendation data set, a past data set, and a current driving data set 1220 .

13 is a visualization of the SAF facility at Unit 1. The entire high-dimensional data set 1310 is visualized as a recommendation data set, a past data set, and a current driving data set 1320 .

14 is a diagram illustrating size selection using a distribution of a training data set according to an embodiment of the present invention. Referring to FIG. 14 , even for the recommended recommended data set, a data size 1410 is recommended from the system by distinguishing a dense portion of data according to a shape in which the data is distributed, or a size is recommended using the distance from the current driving data set. . Also, it may be configured so that a user can select an arbitrary data set 1420 . That is, without selecting the size of the training data, the user can directly set the area using a drag-and-drop method to select the size.

15 is a diagram illustrating size selection using a distance from the center of a training data set according to an embodiment of the present invention. Referring to FIG. 15 , in the training data set 1510 , size selection 1520 using a distance from the center of the training data set is possible.

16 is a diagram illustrating information provision for model correction according to an embodiment of the present invention. Referring to FIG. 16 , by tracing the center of the learning model, it is easy to understand the change in the state of the equipment, and the distance between the center of the learning data and the center of the current state becomes more than a certain distance, and the more alarms occur, the more the model is re-corrected. You can inform the user that this is a necessary situation.

17 is a diagram showing tracking of a learning data pattern according to an embodiment of the present invention. Referring to FIG. 17 , when patterns of learning data are accumulated over a long period of time, it is possible to construct a sophisticated and optimized learning data set over the entire cycle, thereby improving predictive model performance as well as preventing false alarms in advance. can do.

18 is a diagram illustrating occurrence of false positives due to a static threshold according to an embodiment of the present invention. Referring to FIG. 18 , if false alarms were prevented by optimizing the training data of the model previously, in post-processing, a triggering function that does not generate an alarm under the static threshold condition of the residual is combined to limit the predictive model. can be further supplemented. That is, no alarm is generated above or below the static threshold.

19 is a diagram illustrating occurrence of a false detection due to a variable threshold according to an embodiment of the present invention. Referring to FIG. 19 , if false alarms were prevented by optimizing the training data of the model previously, in post-processing, a triggering function that does not generate an alarm under the variable threshold condition of the residual is combined to limit the predictive model. can be further supplemented.

20 is a graph comparing before and after optimization according to an embodiment of the present invention. Referring to FIG. 20 , the sensor value frequently occurs before optimization (2010), but the frequency of the sensor value decreases after optimization (2020). In addition, the reliability before optimization (2010) is 83.5002%, whereas after optimization (2020), the reliability is 97.2979%.

In addition, the steps of the method or algorithm described in relation to the embodiments disclosed herein are implemented in the form of program instructions that can be executed through various computer means such as a microprocessor, a processor, a CPU (Central Processing Unit), etc. It can be recorded on any available medium. The computer-readable medium may include program (instructions) codes, data files, data structures, etc. alone or in combination.

The program (instructions) code recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs, DVDs, Blu-rays, and the like, and ROM, RAM ( A semiconductor memory device specially configured to store and execute program (instruction) code such as RAM), flash memory, and the like may be included.

Here, examples of the program (instruction) code include not only machine language codes such as those generated by a compiler but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

100: Early warning system to prevent false positives
110: communication department
120: collection unit
130: calculator
140: prediction unit
150: output unit
160: storage

Claims (20)

a collection unit 120 for collecting data of power generation facilities;
Calculating data for each signal using the data of the power generation facility, setting the final learning section by performing dimension reduction and data visualization through a recommended learning section recommended using the data for each signal, and updating the target model to update the model a calculation unit 130 that executes model re-learning to generate ; and
a prediction unit 140 for generating prediction information for real-time prediction of the power generation facility in order to prevent false detection by using the update model;
Early warning system for the prevention of false positives comprising a.
The method of claim 1,
The early warning system for preventing false detection, characterized in that the data for each signal is an average value of current operation data for each signal for the data of the power generation facility and an average value of historical data for each signal of the selected past inquiry period.
3. The method of claim 2,
The early warning system for preventing false detection, characterized in that the similarity with the average value of the past data for each signal within the past inquiry period is calculated based on the average value of the current driving data for each signal. 4. The method of claim 3,
The early warning system for the prevention of false positives, characterized in that the similarity is calculated using a distance scale including the Euclidean distance, the Maharanovis distance, and the Manhattan distance.
5. The method of claim 4,
The recommended learning interval is calculated using the final recommendation period determined when the month coincides in two or more distance scales among the distance scales through voting for the month with the highest similarity according to the ranking of the similarity Early warning system for detection prevention.
6. The method of claim 5,
The dimensional reduction is an early warning system for preventing false detection, characterized in that the data for each signal is reduced to a lower dimension and converted into a multivariate dimension of a three-dimensional space.
7. The method of claim 6,
The dimensional reduction is a technique for finding an axis that minimizes the mean square distance between a high-dimensional data set and a projected data set so that the variance of the data for each signal is maximized, or by learning the distribution of the data for each signal to separate the high-dimensional data sets An early warning system for preventing false positives, characterized in that it is performed using a technique of projecting data according to a decision boundary that optimizes .
8. The method of claim 7,
The high-dimensional data set includes an entire data set for a period arbitrarily inquired by a user, a recommended data set recommended from average value-based similarity measurement, and a current operating data set for the last changed state of the power generation facility, characterized in that it includes Early warning system to prevent false positives.
9. The method of claim 8,
The size of the high-dimensional data set is determined by finding the center of the recommended data set and using kurtosis, which is a measure indicating the sharpness of the probability distribution from the center, from which the range of data with high cluster density is calculated from the center. An early warning system for the prevention of false positives, characterized in that it is extracted into a data range of a form close to a normal distribution.
9. The method of claim 8,
The early warning system for preventing false detection, characterized in that the size of the high-dimensional data set is selected through a user's drag-and-drop method.
The method of claim 1,
An early warning system for preventing false detection, characterized in that a variable threshold value with a variable value is set in order to prevent a false alarm that may be caused by a signal pattern that may not be learned in the recommended learning section.
The method of claim 1,
An early warning system for preventing false detection, characterized in that a static threshold value is set to prevent false alarms that may be caused by signal patterns that may not be learned in the recommended learning section.
(a) the collecting unit 120 collecting the data of the power generation facility;
(b) the calculation unit 130 calculates data for each signal using the data of the power generation facility, and performs dimension reduction and data visualization through a recommended learning section recommended using the data for each signal to set the final learning section and executing model retraining to update the target model to generate an updated model; and
(c) generating, by the prediction unit 140, prediction information for real-time prediction of the power generation facility to prevent false detection by using the update model;
An early warning method for preventing false positives comprising a.
14. The method of claim 13,
The early warning method for preventing false detection, characterized in that the data for each signal is an average value of current operation data for each signal for the data of the power generation facility and an average value of historical data for each signal of the selected past inquiry period.
15. The method of claim 14,
The early warning method for preventing false detection, characterized in that the similarity with the average value of the past data for each signal within the past inquiry period is calculated based on the average value of the current driving data for each signal.
16. The method of claim 15,
The early warning method for preventing false positives, characterized in that the similarity is calculated using a distance scale including a Euclidean distance, a Maharanovis distance, and a Manhattan distance. 17. The method of claim 16,
The recommended learning interval is calculated using the final recommendation period determined when the month coincides in two or more distance scales among the distance scales through voting for the month with the highest similarity according to the ranking of the similarity Early warning method to prevent detection.
18. The method of claim 17,
The dimensional reduction is an early warning method for preventing false detection, characterized in that the data for each signal is dimensionally reduced to a lower dimension and converted into a multivariate dimension of a three-dimensional space.
19. The method of claim 18,
The dimensional reduction is a technique for finding an axis that minimizes the mean square distance between a high-dimensional data set and a projected data set so that the variance of the data for each signal is maximized, or by learning the distribution of the data for each signal to separate the high-dimensional data sets An early warning method for preventing false positives, characterized in that it is performed using a technique of projecting data according to a decision boundary that optimizes .
14. The method of claim 13,
The early warning method for preventing false detection, characterized in that a variable threshold value with a variable value is set in order to prevent a false alarm that may be caused by a signal pattern that may not be learned in the recommended learning section.

Images