Device for determining regular test accuracy through signal generation pattern determination by learning

Abstract

The present invention relates to a device for determining regular test accuracy through signal generation pattern determination by learning and, more specifically, to a device for determining regular test accuracy through signal generation pattern determination by learning, whereby it is determined whether a regular test is being properly performed. The present invention comprises: a monitoring unit for monitoring, in real time, an operation limit condition set according to operation technology guidelines for a power plant; an alarm pattern monitoring unit for monitoring alarms that are generated in real time during a regular test, and determining an alarm pattern including an alarm generation order and an alarm generation term; a determination unit (300) for analyzing the alarm pattern for collected alarm data through a learning model to determine a normal alarm pattern and an abnormal alarm pattern generated in a situation deviating from the operation limit condition during the regular test; and a control unit (400) for providing a notification to a user terminal when the abnormal alarm pattern generated in the situation deviating from the operation limit condition during the regular test is detected on the basis of an alarm data analysis result of the determination unit. Therefore, the present invention has the effect of contributing to the safe operation of a power plant by analyzing what information an alarm generated during a regular test contains and providing determination information based on data on whether the regular test is being properly performed.

Description

A device for determining the accuracy of regular tests by identifying signal generation patterns through learning.

[0001] The present invention relates to a device for determining the accuracy of a periodic test through signal generation pattern determination by learning, and relates to a device for determining the accuracy of a periodic test through signal generation pattern determination by learning to confirm whether a periodic test is being properly conducted.

[0002] Power plants ensure safe operation during their operation by complying with the operating restrictions in the operating technical manual.

[0003] These operating restrictions are verified through a procedure called 'periodic' testing to ensure that the equipment or facility in question can perform its inherent safety functions.

[0004] However, although the regular test procedures performed by the maintenance department are very important, there is a problem in that the main control room operator cannot check whether the regular tests are being performed properly according to the procedures.

[0005] Furthermore, due to the multiple channels and procedures involved in performing routine testing procedures, there is a risk of human error that testers may overlook. This can lead to situations where the driver is unable to recognize that the driving restrictions are not being met, potentially leading to violations of the restrictions.

[0006] During regular testing, multiple alarms may be generated by the alarm system, but there is no system that can interpret the information contained in these multiple alarms, and thus, technological development is needed.

[0007] [Prior Art Literature]

[0008] [Patent Document]

[0009] (Patent Document 1) Korean Patent Publication No. 2023-0036178 (March 14, 2023)

[0010] The present invention is intended to solve the above-described problem, and the purpose of the present invention is to provide a device for determining the accuracy of a periodic test through signal generation pattern determination by learning in order to analyze what information an alarm generated during a periodic test contains and to confirm whether the periodic test is being properly performed.

[0011] The present invention includes a monitoring unit (100) that monitors in real time the operating restriction conditions set according to the operating technology guidelines of a power plant, an alarm pattern monitoring unit (200) that monitors alarms that occur in real time during a regular test to confirm an alarm pattern including the alarm occurrence order and alarm occurrence term time, a determination unit (300) that analyzes alarm patterns for alarm data collected through the alarm pattern monitoring unit through a learning model to determine normal alarm patterns and abnormal alarm patterns that occur in a situation where the operating restriction conditions are exceeded during a regular test, and a control unit (400) that provides a notification to a user terminal when an abnormal alarm pattern that occurs in a situation where the operating restriction conditions are exceeded during a regular test is detected based on the alarm data analysis result of the determination unit.

[0012] The alarm pattern monitoring unit (200) includes an alarm receiving module (210) that receives alarm data in real time from the monitoring unit (100), an alarm recording module (220) that records all alarms that have occurred, stores them in a database, and records the occurrence time, type, order, and interval of each alarm, and an alarm pattern analysis module (230) that analyzes alarm data recorded during a regular test in real time to identify an alarm pattern including the occurrence order and interval of alarms.

[0013] The determination unit (300) includes a data collection module (310) that collects alarm data collected through an alarm pattern monitoring unit, a learning data module (320) that uses alarm data collected in a normal driving state and alarm data collected in an abnormal situation as learning data, a learning module (330) that selects a suitable machine learning model to determine an alarm pattern and learns the learning model using the learning data of the learning data module, and a determination module (350) that inputs alarm data coming in in real time into the learning model to determine a normal alarm pattern and an abnormal alarm pattern, and analyzes the alarm pattern through the learning model to determine a normal alarm pattern and an abnormal alarm pattern that occurs in a situation where the driving restriction conditions are exceeded during a regular test.

[0014] The determination unit (300) further includes an evaluation module (340) that evaluates the performance of the learning model using test data, and includes test data used to evaluate the performance of the machine learning model by separating some of the data collected through the alarm pattern monitoring unit (200) from the learning data.

[0015] The determination unit (300) selects a nonlinear learning model to determine an alert pattern, uses the learning data of the learning data module, and learns the learning model to determine normal and abnormal patterns.

[0016] According to the present invention, there is an effect of contributing to the safe operation of a power plant by analyzing what information is contained in alarms generated during regular tests and providing data-based judgment information to determine whether regular tests are being properly performed.

[0017] Figure 1 illustrates the configuration of a device for determining the accuracy of a regular test through signal generation pattern determination by learning according to one embodiment of the present invention.

[0018] FIG. 2 illustrates an alarm pattern monitoring unit of a device for determining the accuracy of a regular test through signal generation pattern determination by learning according to one embodiment of the present invention.

[0019] Figure 3 illustrates a determination unit of a device for determining the accuracy of a regular test through determination of a signal generation pattern through learning according to one embodiment of the present invention.

[0020] The device for determining the accuracy of a regular test by determining a signal generation pattern through learning according to this embodiment is intended to analyze the order and occurrence time interval of alarm patterns that occur when the operating restrictions set according to the operating technology guidelines of a power plant are exceeded, to confirm whether a regular test is being properly performed.

[0021] Accordingly, during regular testing, the alarm system may generate multiple alarms, but the purpose is to interpret the information contained in these alarms. For example, if an abnormal pattern is detected by comparing the sequence or time interval of alarm occurrences with the normal alarm pattern, a notification such as "A cooling system alarm occurred 10 minutes after the turbine temperature alarm. Cooling system inspection required" is sent to the user terminal.

[0022] In this way, by analyzing the numerous alarms that occur during regular tests and determining normal and abnormal patterns based on the order and interval of alarm occurrence, the accuracy of regular tests can be evaluated and abnormal situations can be detected early.

[0023] Driving restrictions can actually be breached during a regular test. A regular test may be a normal test, even if the driving restrictions are breached. If the regular test is performed properly, the alarm trigger pattern should be consistent. For example, if the alarm sequence is consistent, the alarm interval is consistent, and if an abnormality occurs during a regular test, the system does not detect and determine the abnormality. Rather, if the driving restrictions are breached, but the regular test sequence is incorrect or the interval is long, an alarm is generated on the user terminal for the regular test driver.

[0024] According to one embodiment of the present invention, there is provided a device for determining the accuracy of a periodic test through signal generation pattern determination by learning, which analyzes what information an alarm generated during a periodic test contains and confirms whether the periodic test is being properly performed.

[0025] For reference, each alert generated during a regular test contains different information, which can be used to determine the accuracy of the test. By analyzing the cause, time, sequence, and location of the alerts, normal and abnormal patterns can be distinguished. This allows for verification that the test is being conducted properly and prompt response when problems arise.

[0026] The device for determining the accuracy of a regular test through the determination of a signal generation pattern by learning according to this embodiment monitors abnormal situations of each operating restriction condition set according to the operating technology guidelines while the power plant is operating, generates an alarm signal accordingly, learns the alarm signal for each abnormal situation that has occurred, determines the difference between the alarm pattern that has occurred in real time and the normal alarm pattern, and determines whether the regular test is being performed normally through this.

[0027] Hereinafter, the present invention will be described in more detail with reference to the attached drawings.

[0028] Figure 1 illustrates the overall configuration of a device for determining the accuracy of a regular test through signal generation pattern determination by learning according to one embodiment of the present invention.

[0029] As illustrated in Fig. 1, a device (10) for determining the accuracy of a regular test through signal generation pattern determination by learning includes a monitoring unit (100), an alarm pattern monitoring unit (200), a determination unit (300), and a control unit (400).

[0030] The monitoring unit (100) is configured to monitor in real time the operating restriction conditions set according to the operating technology guidelines of the power plant.

[0031] These monitoring units continuously monitor critical operating parameters, such as reactor coolant temperature, pressure, and flow rate, to determine whether operating limits are being violated. If any abnormalities are detected, an alarm is immediately generated.

[0032] For example, if the reactor coolant temperature is to be maintained between 280°C and 320°C, the monitoring unit continuously monitors this. If the temperature rises to 325°C, it is detected and an alarm is immediately generated.

[0033] For reference, the Operational Limiting Conditions (OLC) established in accordance with the operating technical guidelines of a power plant refer to the limit values ​​of specific operating parameters that must be observed to ensure the safety and stable operation of the power plant.

[0034] Operating conditions are established to ensure the normal operation of various devices and systems and to maintain a safe operating environment. While specific operating conditions may vary depending on the type and design of the power plant, they generally include the following elements:

[0035] The reactor coolant temperature must remain within a certain range for the safe operation of the reactor. For example, the reactor coolant temperature must be maintained between 280°C and 320°C.

[0036] The pressure inside the reactor must remain within certain limits for safe operation. For example, the pressure inside the reactor must be maintained between 1,500 and 2,500 psi.

[0037] The coolant flow rate must be above a certain level to ensure sufficient cooling. For example, the coolant flow rate should be at least 5,000 liters / minute.

[0038] The temperature of the reactor fuel must be maintained within safe limits. For example, the fuel temperature must not exceed 1000 degrees Celsius.

[0039] Radiation levels inside and outside the reactor must not exceed acceptable levels, not exceeding 5 millisieverts (mSv) per hour.

[0040] Regarding the operating status of the cooling system, the operating status of the cooling pump, heat exchanger, etc. must be normal, and the operating pressure and flow rate of the pump must be within the specified range.

[0041] At this time, with regard to the operating status of the cooling system, the operating status of the cooling pump and heat exchanger, etc. must be normal, which means that the performance of the cooling system must be maintained stably and efficiently.

[0042] The specific values ​​may vary depending on the design and operating manual of the power plant, but for example, the pressure of the cooling pumps represents the condition that the cooling water must be circulated at an appropriate pressure, which is between 1500 psi (pounds per square inch) and 2500 psi. If the pressure is too low, the cooling water may not be circulated sufficiently, which may prevent proper reactor cooling, and if the pressure is too high, the system may be overloaded and may be damaged.

[0043] The flow rate of a cooling pump refers to the amount of coolant the pump circulates over a given period of time. This directly affects the heat removal efficiency of the cooling system. If the flow rate exceeds the specified range, the reactor's heat may not be sufficiently removed.

[0044] For example, a flow rate between 5,000 and 10,000 liters / min represents sufficient cooling water circulation to effectively remove heat from the reactor. Too low a flow rate could insufficiently cool the reactor, potentially causing the reactor temperature to rise. Too high a flow rate could overload the system, reducing efficiency.

[0045] To give a specific example of operation, the monitoring unit can be considered to be operating stably as the measured value of the pump operating pressure is within the normal range of 1500 psi to 2500 psi while operating at a pressure of 2000 psi.

[0046] If the pump flow rate is within the normal range of 5,000 liters / minute to 10,000 liters / minute, assuming that the pump is circulating the cooling water at a flow rate of 7,000 liters / minute, the cooling system can be considered to be operating efficiently since the pump flow rate is within the normal range.

[0047] Regarding safety system operation, systems for safely shutting down the reactor in an emergency must function normally. Emergency cooling systems, radiation shielding systems, and other systems must function normally, and their operating conditions must remain within specified limits.

[0048] The Emergency Core Cooling System (ECCS) is designed to prevent reactor overheating and minimize fuel damage. The ECCS must have a flow rate of at least 10,000 liters / minute. The ECCS pressure must be between 1,500 and 2,500 psi, and the coolant temperature must be between 50 and 150°C.

[0049] Radiation shielding systems are designed to maintain radiation levels within and around power plants at safe levels. Radiation levels must not exceed 5 millisieverts (mSv) per hour, and the materials used for radiation shielding must be in good condition. The thickness, density, and other characteristics of the shielding materials must meet established standards.

[0050] Simultaneously monitor the operating conditions of the safety system's emergency cooling system and radiation shielding system, so that an alarm can be generated if they exceed the specified range.

[0051] Regarding chemical concentrations, the chemical composition of coolants and other operating materials must be maintained within specified concentrations. The boron concentration in coolants must be maintained between 1,000 and 2,000 ppm.

[0052] The alarm pattern monitoring unit (200) monitors alarms that occur in real time while a regular test is in progress and checks the alarm pattern including the alarm occurrence order and alarm occurrence term time.

[0053] That is, the alarm pattern monitoring unit (200) monitors all alarms occurring in real time during regular testing. It records various alarms occurring during testing in real time and monitors the patterns of these alarms. The data collected during this process is used for subsequent analysis.

[0054] For example, the alarm pattern monitoring unit (200) records alarms occurring during a performance test of a reactor coolant circulation pump. For example, if a series of alarms occur due to a temperature rise, all of them are recorded and used for subsequent analysis.

[0055] Periodic testing refers to a series of procedures regularly performed to maintain the safety and reliability of a power plant. These tests are conducted to ensure that various equipment and systems at the power plant are functioning normally and comply with established operating conditions.

[0056] Examples include reactor cooling system tests to ensure that the cooling system operates normally and effectively removes heat from the reactor; emergency cooling system tests to ensure that the reactor can be safely cooled in an emergency; and radiation shielding system tests to ensure that the radiation shielding system operates normally and prevents radiation leaks.

[0057] The alarm pattern monitoring unit (200) according to this embodiment detects and records all alarms occurring during a regular test in real time, and receives and monitors alarms occurring in the monitoring unit (100) in real time.

[0058] Regarding alarm data recording, the time of occurrence, type, severity, etc. of each alarm are recorded. Additionally, all consecutive alarms are recorded without omission. The recorded data is stored in a database for subsequent analysis. If a specific alarm occurs unusually frequently or continuously, this can be detected and transmitted to the determination unit (300).

[0059] FIG. 2 illustrates an alarm pattern monitoring unit of a device for determining the accuracy of a regular test through signal generation pattern determination by learning according to one embodiment of the present invention.

[0060] This alarm pattern monitoring unit (200) includes an alarm receiving module (210) that receives alarm data in real time from the monitoring unit (100), an alarm recording module (220) that records all alarms that have occurred, stores them in a database, and systematically records the time of occurrence, type, severity, etc. of each alarm, and an alarm pattern analysis module (230) that analyzes alarm data recorded during a regular test in real time to identify an alarm pattern including the order and interval of alarm occurrence.

[0061] The alarm recording module (220) systematically records the time, type, order, and interval of occurrence of each alarm. Each time an alarm occurs, the time is recorded in timestamp format. For example, if Alarm A occurs at 10:00:00 on 2024-06-28, this exact time is recorded.

[0062] Additionally, the types of alerts are recorded by classifying them into predefined categories. For example, if alert A is type 1, it is recorded as "Type 1." Additionally, alarms are assigned sequential numbers based on the order in which they occurred. The first alarm is recorded as sequence 1, the second as sequence 2, etc. Furthermore, the interval between each alert and the previous alert is calculated and recorded. For example, if the interval between alert A and alert B is 10 seconds, this is recorded as "10 seconds."

[0063] The alarm pattern analysis module (230) according to this embodiment can analyze alarm data recorded during a regular test in real time, check the alarm occurrence order and occurrence interval, and compare it with a normal pattern.

[0064] The alarm pattern monitoring unit (200) is configured to monitor alarms occurring in real time while a regular test is in progress and check the alarm occurrence order and alarm occurrence interval, and the alarm pattern analysis module (230) analyzes data in real time to monitor the alarm occurrence order and alarm occurrence interval.

[0065] The determination unit (300) learns normal alarm patterns and abnormal alarm patterns for alarm patterns including the alarm occurrence order and alarm occurrence term time from the alarm data collected through the alarm pattern monitoring unit through a learning model, and determines normal alarm patterns and abnormal alarm patterns in situations where the driving restriction conditions are exceeded during a regular test.

[0066] That is, the determination unit (300) according to the present embodiment analyzes the alarm pattern for the alarm data collected through the alarm pattern monitoring unit through a learning model to determine the normal alarm pattern and the abnormal alarm pattern that occur in a situation where the driving restriction conditions are exceeded during a regular test.

[0067] This discriminator uses a machine learning model to distinguish between normal and abnormal alert patterns based on collected alert data. The learning model is trained using historical data, effectively identifying even nonlinear patterns.

[0068] Here, the learning model is a machine learning model that learns past data to distinguish between normal and abnormal alert patterns.

[0069] The detection unit learns from the collected data patterns in which the reactor coolant temperature deviates from the normal range for a certain period of time and identifies these as abnormal. For example, if the coolant temperature persists above 325°C, this is recognized as an abnormal pattern.

[0070] In addition, the judgment unit analyzes the alarm pattern of the alarm data collected through the alarm pattern monitoring unit through a learning model, and detects abnormal patterns by analyzing the alarm pattern including the alarm occurrence order and alarm occurrence term time (interval) that occur in situations where the driving restriction conditions are exceeded in the data collected during the regular test.

[0071] Figure 3 illustrates a determination unit of a device for determining the accuracy of a regular test through determination of a signal generation pattern through learning according to one embodiment of the present invention.

[0072] The determination unit of the device for determining the accuracy of regular testing through signal generation pattern determination by learning according to this embodiment can learn and determine various alarm patterns by utilizing a machine learning model. This determination unit (300) includes a data collection module (310), a learning data module (320), a learning module (330), an evaluation module (340), and a determination module (350), thereby effectively distinguishing between normal and abnormal patterns, thereby enabling the regular testing procedures of a power plant to be performed accurately.

[0073] The data collection module (310) collects alarm data collected through the alarm pattern monitoring unit.

[0074] For example, among the alarm data collected through the alarm pattern monitoring unit, data such as reactor coolant temperature, pressure, and flow rate are collected. At this time, the collected data is converted into a format suitable for analysis. This process includes data normalization, missing value processing, and outlier removal.

[0075] The learning data module (320) prepares learning data by using alarm data collected in normal driving conditions as normal pattern data and alarm data collected in abnormal situations as abnormal pattern data.

[0076] The learning module (330) selects an appropriate machine learning model to determine an alert pattern and trains the model using the learning data from the learning data module. For example, a Random Forest Classifier can be used.

[0077] The evaluation module (340) uses test data to evaluate the performance of the learning model. This includes evaluation indicators such as accuracy, precision, and recall. This evaluation module calculates indicators for evaluating the model's performance using the TP (True Positive), TN (True Negative), FP (False Positive), and FN (False Negative) values, thereby enabling the evaluation of the learning model's performance.

[0078] TP (True Positive) is the number of cases where the learning model predicted a normal pattern and the sample was actually a normal pattern, and it is the number of cases where the model predicted a normal pattern in the alert data and the pattern was actually a normal pattern.

[0079] TN (True Negative) is the number of cases where the learning model predicted an abnormal pattern and the pattern was actually abnormal, and it is the number of cases where the model predicted an abnormal pattern in the alert data and the pattern was actually abnormal.

[0080] False Positive (FP) is the number of instances where a learning model predicts a normal pattern but actually detects an abnormal pattern. For example, in alert data, the model predicts a normal pattern but it is actually an abnormal pattern. This is when the model incorrectly predicts "normal."

[0081] False Negative (FN) is the number of instances where a learning model predicts an abnormal pattern but actually turns out to be a normal pattern. For example, in alert data, the model predicts an abnormal pattern but it is actually a normal pattern. This is when the model incorrectly predicts an "abnormal" pattern.

[0082] Accuracy indicates how accurately the learning model predicts across all samples. It is calculated as (TP + TN) / (TP + TN + FP + FN).

[0083] Precision is the proportion of samples predicted by the learning model to be normal patterns that actually have a normal pattern. This indicates the accuracy of the model's predictions. It is calculated as TP / (TP + FP).

[0084] Recall represents the percentage of normal patterns correctly predicted by the learning model among actual normal pattern samples. This represents the model's sensitivity. It is calculated as TP / (TP + FN).

[0085] Test data is a dataset used to evaluate the performance of a machine learning model. Test data is separate data not used in model training and is used to verify how well the model performs in real-world environments.

[0086] Test data can be constructed by separating some of the data collected through the alarm pattern monitoring unit (200) from the training data. For example, various alarm data generated during a performance test of a reactor coolant circulation pump can be used as test data.

[0087] The judgment module (350) inputs real-time incoming alarm data into a learning model to determine normal alarm patterns and abnormal alarm patterns.

[0088] Note that while driving restrictions may actually be exceeded during a regular test, they may still be normal. While warnings may be issued during a regular test that deviate from the expected driving restrictions, if the regular test is conducted properly, the warning pattern should remain the same.

[0089] For example, normal alert patterns should be present, such as one alert being triggered followed by another, the same order of alerts, or the same alert term (interval) time between alerts.

[0090] In this embodiment, the occurrence of an alarm does not mean that an abnormal situation occurs during a regular test and an alarm is generated by judging it as an abnormal situation, but that an operating restriction condition may be exceeded during a regular test, but if the test order of the regular test is incorrect, or if the order in which an alarm should be generated is changed, or if the alarm generation interval is too long or too short, an alarm corresponding to an abnormal alarm pattern is generated to the user terminal.

[0091] The device for determining the accuracy of a periodic test by determining a signal generation pattern through learning according to this embodiment may exceed the operating restriction conditions when conducting a periodic test, so the driver can recognize whether the alarm is normal even if it sounds during a periodic test.

[0092] That is, what the device for determining the accuracy of a regular test through the determination of a signal generation pattern through learning detects is when the occurrence pattern of an alarm that deviates from the operating restriction conditions that occurs during a regular test is different.

[0093] The discriminator of the device for determining the accuracy of regular testing through learning-based signal generation pattern discrimination according to this embodiment uses an artificial intelligence model capable of effectively identifying nonlinear patterns. Models capable of nonlinear discriminant expressions have the ability to learn and predict complex data patterns beyond simple linear relationships. Using these models, normal and abnormal patterns can be more accurately distinguished from various alarm data from power plants.

[0094] Nonlinear patterns refer to situations where the relationships between data variables are not simple linear relationships. Various alarm data generated from power plants can be nonlinear for the following reasons: The interactions between various sensor data (temperature, pressure, flow rate, etc.) are complex and cannot be explained by a simple linear model. Furthermore, nonlinear patterns can emerge when abrupt changes in system status or abnormalities occur. The complex interplay of various variables, such as environmental conditions, equipment status, and operating conditions, exhibits nonlinear characteristics.

[0095] Examples of machine learning models that are useful for identifying alert patterns using nonlinear discriminants include:

[0096] Random Forests improve predictive performance by combining multiple decision trees, each capable of performing nonlinear classification. In other words, they address nonlinearity by synthesizing the prediction results from multiple trees.

[0097] Support vector machines (SVMs) are models that find boundaries that classify data in high-dimensional space. They can learn nonlinear patterns using the kernel trick, demonstrating excellent performance in nonlinear classification problems.

[0098] An artificial neural network (ANN) is a neural network model that mimics the structure of the human brain. It uses multiple layers to learn complex patterns. Deep learning allows it to learn extremely complex nonlinear relationships.

[0099] Note that multiple alarms generated during regular testing may be intentionally triggered to verify the system's normal operation. This process may generate multiple alarms, and analyzing the pattern of these alarms can help assess the accuracy of regular testing.

[0100] An abnormal alarm pattern during a regular test refers to an alarm pattern that deviates from the normal sequence and interval of alarm occurrences that occur during a regular test when operating restrictions are exceeded. Alarms occurring during a regular test can be analyzed by comparing them with patterns observed during normal operation.

[0101] This allows the judgement unit to define abnormal patterns as those that deviate from the sequence and interval of alarms that should occur during normal routine testing.

[0102] A normal alarm pattern is the expected sequence and interval of alarms that occur during a normal routine test. For example, if a normal alarm pattern is "turbine temperature alarm - cooling system alarm" occurring at 5-minute intervals, this is a learned pattern based on past normal routine test data.

[0103] An abnormal alarm pattern is one in which alarms occur in a different order or interval than would be expected during a normal routine test, for example, a "turbine temperature alarm" should occur 5 minutes after a "cooling system alarm" in a normal alarm pattern, but in fact occurs 10 minutes later, or a "boiler pressure alarm" occurs instead.

[0104] In pattern identification, the collected alarm data is compared with alarm data from past normal regular tests, and the alarm pattern is analyzed through a learning model. If it deviates from the normal alarm occurrence sequence and interval, it is identified as an abnormal pattern.

[0105] In other words, a normal pattern is when a "turbine temperature alarm" occurs during a normal periodic test and then a "cooling system alarm" occurs 5 minutes later, and an abnormal pattern is when a "cooling system alarm" occurs 10 minutes later instead of 5 minutes after a "turbine temperature alarm" occurs, or when a "boiler pressure alarm" occurs instead of a "cooling system alarm." If the alarms occur in the order of "turbine temperature alarm → boiler pressure alarm" or the interval is excessively long or short, this is determined to be an abnormal pattern.

[0106] The judgment unit analyzes the collected alarm data through a learning model to determine normal and abnormal patterns, and identifies abnormal patterns by comparing the order and interval of alarm occurrences that exceed operating restrictions during regular testing.

[0107] The control unit (400) immediately provides a notification to the user terminal when an abnormal pattern is detected based on the analysis results of the alarm data of the determination unit.

[0108] For example, a notification such as 'Cooling system alarm occurs 10 minutes after turbine temperature alarm exceeding the expected interval (5 minutes) and an abnormal pattern is detected, so cooling system inspection is required' is sent.

[0109] These control units immediately send notifications to the operator or user terminal (e.g., a relevant personnel terminal) when abnormal patterns are detected. This allows the operator to respond quickly and ensures that regular testing procedures are performed accurately.

[0110] For example, if an alarm identified as an abnormal pattern is detected by the judgment unit, the control unit sends a notification to the operator's terminal stating, "The test procedure is being performed abnormally due to an excessive coolant temperature." The operator acknowledges this notification and takes immediate action to resolve the issue.

[0111] The operation of the device for determining the accuracy of a regular test through signal generation pattern determination by learning according to this embodiment is described as follows.

[0112] The control unit (400) monitors in real time the operating restriction conditions set according to the power plant's operating technology guidelines through the monitoring unit (100). Thereafter, the control unit (400) monitors all alarms occurring in real time during the periodic test through the alarm pattern monitoring unit (200). The control unit (400) analyzes the alarm data collected from the alarm pattern monitoring unit through the determination unit (300) and determines normal and abnormal patterns through a learning model.

[0113] According to the present invention, there is an effect of contributing to the safe operation of a power plant by analyzing what information is contained in alarms generated during regular tests and providing data-based judgment information to determine whether regular tests are being properly performed.

Claims

1. A monitoring unit (100) that monitors in real time the operating restrictions set according to the operating technology guidelines of the power plant. An alarm pattern monitoring unit (200) that monitors alarms occurring in real time while a regular test is in progress and checks the alarm pattern including the alarm occurrence order and alarm occurrence term time. A determination unit (300) that analyzes the alarm pattern of the alarm data collected through the above alarm pattern monitoring unit through a learning model and determines the normal alarm pattern and abnormal alarm pattern that occur in a situation where the driving restriction conditions are exceeded during a regular test, and A device for determining the accuracy of a regular test through signal generation pattern determination by learning, characterized in that it includes a control unit (400) that provides a notification to a user terminal when detecting an abnormal alarm pattern that occurs in a situation where the driving restriction conditions are exceeded during a regular test based on the results of the analysis of the alarm data of the above determination unit.

2. In paragraph 1, The above alarm pattern monitoring unit (200) An alarm receiving module (210) that receives alarm data in real time from the monitoring unit (100), An alarm recording module (220) that records all alarms that occur, stores them in a database, and records the occurrence time, type, order, and interval of each alarm. A device for determining the accuracy of a periodic test through signal generation pattern determination by learning, characterized by including an alarm pattern analysis module (230) that analyzes alarm data recorded during a periodic test in real time to identify an alarm pattern including the alarm occurrence order and occurrence interval.

3. In paragraph 1, The above determination unit (300) is A data collection module (310) that collects alarm data collected through the alarm pattern monitoring unit; A learning data module (320) that uses alarm data collected in normal driving conditions and alarm data collected in abnormal situations as learning data. A learning module (330) that selects a suitable machine learning model to determine an alert pattern and trains the learning model using the learning data of the learning data module. A device for determining the accuracy of a periodic test through determination of a signal generation pattern by learning, characterized in that it includes a determination module (350) that inputs real-time incoming alarm data into a learning model to determine a normal alarm pattern and an abnormal alarm pattern, analyzes the alarm pattern through the learning model to determine a normal alarm pattern and an abnormal alarm pattern that occurs in a situation where the driving restriction conditions are exceeded during a periodic test.

4. In paragraph 1 or paragraph 3, The above determination unit (300) A device for determining the accuracy of a regular test through determination of a signal generation pattern by learning, characterized in that it further includes an evaluation module (340) for evaluating the performance of a learning model using test data, and includes test data used to evaluate the performance of a machine learning model by separating some of the data collected through the alarm pattern monitoring unit (200) from the learning data.

5. In paragraph 1, The above-mentioned determination unit (300) is a device for determining the accuracy of a regular test through determination of a signal generation pattern by learning, characterized in that it selects a nonlinear learning model to determine an alarm pattern, uses learning data of a learning data module, and learns the learning model to determine a normal pattern and an abnormal pattern.

Images