KR20240032413A - System and operating method for diagnosing plant based on intelligent agent using simulator - Google Patents

Abstract

The present invention relates to an artificial intelligence-based plant diagnosis system using a simulator and its operation method. The simulator is configured to reflect the characteristics of the plant, the multiple models of the simulator are linked and integrated to create an integrated linkage model, and the required Simulation is performed by configuring scenarios corresponding to operating conditions, and a learning model is created to diagnose and predict plant failures based on simulation data to diagnose plant failures and predict abnormalities, so that the actual plant is not operated. Even so, reliable simulation data for various situations can be secured, learning data can be easily secured by reflecting normal and abnormal states, and the accuracy of diagnosis of failures and abnormalities in plants can be improved.

Description

Artificial intelligence-based plant diagnosis system using a simulator and its operation method {SYSTEM AND OPERATING METHOD FOR DIAGNOSING PLANT BASED ON INTELLIGENT AGENT USING SIMULATOR}

The present invention is an artificial intelligence-based plant diagnosis system using a simulator that diagnoses and predicts plant failure by accumulating simulation data according to plant process and control in various environments using a simulator to create a learning model for plant failure diagnosis. and its operation method.

Currently, in power generation plants, a method is used to install sensors to detect abnormalities and failures in equipment and analyze the degree of change in process variables such as temperature, pressure, and flow rate based on the measured signals.

Plant abnormalities are operated and managed through alarms in the plant control system, but because alarms are generated after a failure occurs, it is impossible to predict and prepare for a failure before it occurs, and there is often insufficient time to respond after a failure occurs, causing problems. There is.

Accordingly, recently, techniques for diagnosing and predicting plant abnormalities using plant operation data and artificial intelligence technology have been developed.

Related technologies include Republic of Korea Patent No. 10-2118966, “Power Plant Failure Prediction and Diagnosis System and Method.”

An artificial intelligence model that diagnoses abnormal states of equipment or predicts abnormalities/failures in advance not only requires a large amount of historical data (big data) on well-selected normal and abnormal states, but also participates in the learning of the artificial intelligence model. The reliability of the data to be analyzed has clear limits that can determine the performance of the artificial intelligence model.

Therefore, in order to diagnose and predict plant abnormalities reliably, a vast amount of data on normal and abnormal states generated during plant operation is required. In particular, the data to be trained on the artificial intelligence model is normal data in the stable operating range of the plant and When a plant abnormality occurs, data reflecting the cause of the abnormality and the results of action are needed.

However, due to the nature of power system operation, which is sensitive to power demand and supply, it is difficult to view the operation data generated from plants connected to the grid as steady-state data that can be easily learned by artificial intelligence models, and securing abnormal state data is even more difficult.

Operation in the abnormal area falls within a significantly smaller operating range compared to operation in the normal area, and in order to secure abnormal state operation data, it cannot have abnormal effects on well-operated facilities.

Therefore, a method is needed to secure reliable plant data in various situations and build a learning model according to the corresponding variables or operating conditions to not only diagnose plant failures but also predict failures.

Republic of Korea Patent No. 10-2118966

The present invention was created in response to the above-mentioned need. By configuring a simulator to reflect the characteristics of the plant, configuring a scenario corresponding to the required operating conditions, and performing a simulation, data on a specific situation is secured and plant failure diagnosis is achieved. The purpose is to provide an artificial intelligence-based plant diagnosis system using a simulator and its operation method, which is applied to .

In order to achieve the above object, an artificial intelligence-based plant diagnosis system using a simulator according to the present invention includes a simulator that configures a process model and a control model corresponding to the plant and performs simulation according to operating conditions to generate simulation data; and a diagnostic prediction device that generates a learning model for the plant based on the simulation data, analyzes data of the plant based on the learning model, diagnoses the state of the plant, and predicts abnormalities. It includes, and the simulator is characterized in that it creates an integrated linkage model by linking input and output variables between the process model and the control model, and performs simulation by generating a scenario for each operating condition.

The simulator includes a linkage model including external environmental conditions for the process model and the control model; An operation control panel that outputs progress according to the simulation; and an instructor operation panel that runs the simulator to perform simulation. It further includes.

The simulator is characterized by tuning the parameters of the process model and linking input and output variables by modifying the logic of the control model in response to the tuned parameters.

The simulator integrates and links the analog monitoring variables and face plate variables of the control model with the variables of the input/output model, registers tags in the input/output model, and outputs them on the screen of the operation control panel.

The simulator is characterized in that it generates the integrated linkage model by reflecting simulation operation procedure data based on the operation procedure of the plant.

The simulator is characterized in that it verifies the integrated linkage model by calculating the fidelity of the integrated linkage model by reflecting the design data and operation data of the plant.

The simulator is characterized by maintaining thermal balance at 100% rated load and determining fidelity to the integrated linkage model by checking whether key data satisfies design data in the thermal balance state.

The simulator is characterized by lowering the load to the thermal balance state at 100% load, switching to the cold state, and operating up to the rated load to determine fidelity to the integrated linkage model.

The simulator generates a scenario for each operating condition for at least one of cold state, hot state, gas turbine preheating, startup, heat recovery boiler preheating, steam turbine speed up, grid feed-in, output 50%, output 75%, and output 100%. It is characterized by:

The diagnosis prediction device generates a normal state diagnosis prediction model and an abnormal state diagnosis prediction model based on a learning model generated by learning simulation data generated from the simulator, and determines a normal state based on the steady state diagnosis prediction model. Diagnosing and predicting the state of the plant according to abnormal operating conditions, and diagnosing and predicting failure of the plant due to an abnormal state based on the abnormal state diagnosis and prediction model.

The diagnostic prediction device diagnoses and predicts the state of the plant according to operating conditions that deviate from the normal state based on the normal state diagnostic prediction model, and detects a failure of the plant due to an abnormal state based on the abnormal state diagnostic prediction model. It is characterized by diagnosis and prediction.

A method of operating an artificial intelligence-based plant diagnosis system using a simulator of the present invention includes the steps of configuring, by the simulator, a process model and a control model corresponding to a plant; Generating an integrated linkage model by linking input and output variables between the process model and the control model; Generating simulation data by creating a scenario for each driving condition and performing a simulation; A diagnostic prediction device generating a learning model for the plant based on the simulation data; and analyzing data from the plant based on the learning model to diagnose the state of the plant and predict abnormalities. Includes.

The step of generating the integrated linkage model includes configuring a linkage model including external environmental conditions for the process model and the control model; and integrating the process model, the control model, and the linkage model; generating the integrated linkage model by reflecting simulation operation procedure data based on the operation procedure of the plant; Includes.

Generating the integrated linkage model includes tuning parameters of the process model; Linking input and output variables by modifying the logic of the control model in response to the tuned parameters; Integrating and linking analog monitoring variables and face plate variables of the control model with variables of the input/output model; and registering a tag in the input/output model and outputting it on the operation panel screen of the simulator. Includes.

After generating the integrated linkage model, verifying the integrated linkage model by calculating fidelity to the integrated linkage model by reflecting the design data and operation data of the plant; It further includes.

The step of verifying the integrated linkage model includes setting the simulator to maintain a thermal balance state at a rated 100% load; and determining fidelity to the integrated linkage model by checking whether key data obtained in thermal equilibrium satisfies the design data. Includes.

Verifying the integrated linkage model includes: switching the simulator to a cold state by lowering the load from the thermal balance state at 100% load; and after switching, starting up to the rated load and determining fidelity to the integrated linkage model; It further includes.

The step of diagnosing the state of the plant and predicting an abnormality includes generating a normal state diagnosis prediction model and an abnormal state diagnosis prediction model based on a learning model generated by learning simulation data generated from the simulator; Diagnosing and predicting the state of the plant according to operating conditions outside of the normal state based on the steady-state diagnosis and prediction model; and diagnosing and predicting a failure of the plant due to an abnormal state based on the abnormal state diagnosis and prediction model; Includes.

According to one aspect, the artificial intelligence-based plant diagnosis system using the simulator of the present invention and its operation method configure the simulator by reflecting the operation characteristics of the process and control for the facility system of the plant, and configure a scenario corresponding to the operating conditions. By performing simulations, big data and operating models for various situations can be created even if the actual plant is not operated.

According to one aspect of the present invention, learning data is secured by reflecting normal and abnormal states respectively, and data for a specific situation is obtained by comparing actual data and simulation data, and applied to diagnosis and prediction, thereby detecting failures and abnormalities for the plant. The accuracy of prediction can be improved.

Figure 1 is a diagram illustrating the configuration of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.
Figure 2 is a flow chart of a simulation method of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.
Figure 3 is a flowchart showing a diagnosis method of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.
FIG. 4 is a diagram referenced for explaining the configuration of a simulator of a plant diagnosis system according to an embodiment of the present invention.
Figure 5 is a diagram referenced to explain the simulation process of the plant diagnosis system according to an embodiment of the present invention.
Figure 6 is a diagram referenced to explain scenario configuration and learning model creation of a plant diagnosis system according to an embodiment of the present invention.

Hereinafter, the present invention will be described with reference to the attached drawings.

In this process, the thickness of lines or sizes of components shown in the drawing may be exaggerated for clarity and convenience of explanation. In addition, the terms described below are terms defined in consideration of functions in the present invention, and may vary depending on the intention or custom of the user or operator. Therefore, definitions of these terms should be made based on the content throughout this specification.

Figure 1 is a diagram illustrating the configuration of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.

As shown in FIG. 1, the plant diagnosis system includes a simulator 100, a database (DB) 50, and a diagnosis prediction device 200.

The simulator 100 is a system that can simulate the actual operation of a plant by mathematically simulating the process and control functions of field equipment. The simulator 100 is configured to reflect the process, control, and operation characteristics of the plant's equipment system.

The simulator 100 can secure big data by creating a scenario corresponding to the required operating conditions and performing a simulation on the process/control variables set in response to the scenario.

The simulator 100 constructs normal state data and abnormal state data generated through simulation. The simulator 100 stores the generated data in the database (DB) 50.

The simulator 100 includes at least one operation control panel (110, 110a, 110b), an instructor control panel 120, a control model 130, and a process model 140.

The diagnostic prediction device 200 applies an algorithm to predict abnormalities or failures in the plant based on data generated by the simulator 100, and generates a normal state learning model 20 for the normal state and an abnormal state for the abnormal state. Create a learning model (30).

The diagnostic prediction device 200 analyzes the data of the plant 10 using the normal state learning model 20 and the abnormal state learning model 30, diagnoses existing failures, and provides information on possible abnormalities. Diagnose.

Additionally, the diagnosis and prediction device 200 updates the learning model by reflecting the diagnosis and prediction results based on plant data in the learning model.

The database (DB) 50 stores big data of the simulator 100, and the normal state learning model 20 and abnormal state learning model 30 of the diagnosis and prediction device 200.

Additionally, the database (DB) 50 stores plant data and stores diagnosis results and prediction results based on the plant data.

Figure 2 is a flow chart of a simulation method of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.

As shown in FIG. 2, the simulator 100 configures a process model 140 corresponding to the actual plant 10 and a control model 130 corresponding to the control system of the plant 10.

The instructor operation panel 120 operates the simulator 100 to perform simulation (S10).

The instructor operation panel 120 of the simulator 100 sets variables for simulation of the process model 140 and the control model 130 of the simulator 100 (S21) (S22).

The instructor operation panel 120 monitors the menu configuration and progress for the simulation through the operation operation panel 110 (S23).

In addition, the instructor operation panel 120 configures a linkage model and links data on variables necessary for simulation, operating conditions, and external environmental variables to the simulator 100 (S24).

The instructor operation panel 120 acquires operation procedure data for simulation based on the plant's operation procedures (S31).

The instructor operation panel 120 integrates the model using the simulator's process model 140, control model 130, linkage model, and operation procedure data, and creates a linkage test for it (S35).

The instructor operation panel 120 performs a simulation by applying the plant design and operation data of the actual plant 10 to the simulator 100, and verifies the fidelity of the simulator 100 based on the results (S45).

If the fidelity of the simulator 100 is higher than the set value, the instructor operation panel 120 completes the configuration of the simulator, and if the fidelity is less than the set value, it modifies the simulator 100 by re-reflecting the data of the plant 10. The instructor operation panel (1200) modifies the process model (140), control model (130), and linkage model, and revises the plant operation procedures and design and operation data to modify the simulator (100).

The instructor operation panel 120 creates a scenario based on the required driving conditions (S50). The instructor operation panel 120 generates a scenario for each driving condition.

In addition, the instructor operation panel 120 authorizes failure simulation (S48) and creates a scenario for the failure situation (S50).

The instructor operation panel 120 performs a simulation of the normal state and generates normal relative simulation data (S60). The instructor operation panel 120 performs simulation based on steady-state conditions and operating situations to generate steady-state simulation data for the plant 10.

The instructor operation panel 120 performs a simulation of an abnormal state and generates normal relative simulation data (S65). The instructor operation panel 120 performs simulation based on conditions and operating situations for abnormal states, failures, and non-normal states, and generates normal state simulation data for the plant 10.

The simulation process is output through the operation control panel 110.

The operation control panel 110 pauses or restarts the simulation when a specific key is input during the simulation process.

The instructor operation panel 120 stores the generated normal-state simulation data and abnormal-state simulation data in the database (DB) 50.

The simulator 100 repeats the simulation for each scenario, accumulates the corresponding data, and stores it in the database (DB) 50.

Figure 3 is a flowchart showing a diagnosis method of an artificial intelligence-based plant diagnosis system using a simulator according to an embodiment of the present invention.

As shown in FIG. 3, the steady-state simulation data and abnormal-state simulation data generated by the simulator 100 are respectively stored in the database (DB) 50 (S60) (S65).

The diagnostic prediction device 200 calls data stored in the database (DB) 50 and performs diagnostic prediction.

The diagnosis prediction device 200 learns the normal state data of the plant 10 by applying an abnormality diagnosis prediction algorithm based on the normal state simulation data (S70).

In addition, the diagnosis prediction device 200 learns abnormal state data of the plant 10 by applying an abnormality diagnosis prediction algorithm based on abnormal state simulation data (S75).

The abnormality diagnosis prediction algorithm uses an artificial intelligence-based learning algorithm such as machine learning or deep learning. For example, Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long short term memory network (LSTM), Restricted Boltzmann Machine (RBM), and Deep Belief Network (DBN) may be used.

The diagnostic prediction device 200 configures a learning model by merging the learning results for the normal state data and the learning results for the abnormal state data (S85).

In addition, the diagnosis prediction device 200 performs data labeling on learning results (S81), analyzes correlations between data (S88), and constructs a learning model reflecting the analysis results (S85).

The diagnosis prediction device 200 generates a normal state learning model 20 for normal state diagnosis prediction based on the generated learning model (S90), and an abnormal state learning model 30 for abnormal state diagnosis prediction. Then (S95), each is saved.

The diagnostic prediction device 200 applies and analyzes data from the plant 10 based on the normal state learning model 20, and calculates and provides a diagnostic prediction solution for the state of the plant that is out of normal state (S97). A state outside the normal state is a state in which some abnormality has occurred in the plant before it can be classified as a failure, although it is not normal.

The diagnostic prediction device 200 applies and analyzes data from the plant 10 based on the abnormal state learning model 30, and calculates and provides a diagnostic prediction solution for the abnormal state of the plant 10 (S97). . Abnormal states may include cases where a failure occurs and cases where a failure occurs even though a warning has not occurred.

FIG. 4 is a diagram referenced for explaining the configuration of a simulator of a plant diagnosis system according to an embodiment of the present invention.

The simulator 100 includes at least one operation control panel 110, an instructor control panel 120, a control model 130, and a process model 140.

As shown in (a) of FIG. 4, the process model 140 is constructed by simulating the entire process of the actual plant 10.

As shown in (b) of FIG. 4, the control model 130 is configured by simulating a control system for controlling the process of the actual plant 10.

As shown in (c) of FIG. 4, the operation control panel 110 displays the operation status and operation flow for the process model 140 and the control model 130. The driving control panel 110 may further include a second driving control panel 110a and a third driving control panel 110b.

There is at least one operation control panel 110 and may be composed of multiple units. The operation control panel 110 includes a display capable of displaying simulation situations.

As shown in (d) of FIG. 4, the instructor operation panel 120 operates the simulator 100.

The instructor operation panel 120 starts and stops the simulator 100 to perform simulation. The instructor operation panel 120 creates and stores a scenario, calls a previously created scenario, applies it to the simulation, performs a failure simulation, and performs the simulation by applying external environmental variables to the simulation.

The instructor operation panel 120 performs failure simulation by manipulating the variables necessary for failure simulation in the process model 140 and the control model 130.

The instructor operation panel 120 reflects abnormal phenomena of the equipment in normal state operation through the simulator 100. Accordingly, the simulator 100 can calculate abnormal state data for abnormal conditions or failure conditions corresponding to the reflected phenomenon.

When the failure simulation function of the instructor operation panel 120 is activated, phenomena related to the abnormal state of the plant are simulated. At this time, the operation control panel 110 outputs control and process values that reflect the phenomena accompanying failure simulation.

When the plant type is changed, the simulator 100 can simulate it by changing the process model 140 and control model 130 for the plant type.

Figure 5 is a diagram referenced to explain the simulation process of the plant diagnosis system according to an embodiment of the present invention.

As shown in FIG. 5, the simulator 100 creates a linkage model for input/output variables and external conditions between models of the simulator and performs a model integration linkage test.

Additionally, the simulator 100 can perform a linkage test through the simulator 100 by applying operation procedure data based on the plant operation procedure.

The simulator 100 configures the OPC input/output model 150 between the control model 130 and the operation panel 110, and configures the input/output model 160 between the control model 130 and the process model 140. , integrate input and output variables between models, and connect multiple models to perform linkage tests.

The simulator 100 outputs the progress and abnormal status of the linkage test on the screen 170 of the operation control panel.

The simulator 100 tunes parameters within the process model and modifies the logic of the control model 130 in response to the tuned parameters.

The simulator 100 integrates and links the analog monitoring variables and face plate variables of the control model 130 with the variables of the OPC input/output model 150.

Additionally, the simulator 100 links analog monitoring variables and face plate variables with the process model 140 through the input/output model 160.

For example, the input/output model 160 connected between the control model 130 and the process model 140 includes analog monitoring variables such as COP Min Flow (CM), COP Min Flow (PM), Hotwell Level (CM), and Hotwell Level ( PM), and the face plate variable can be linked to COP Min FCV(CM) and COP Min FCV(PM).

In addition, the OPC input/output model 150 connected between the control model 130 and the operation panel 110 is linked to COP Min Flow (CM), COP Min Flow (HMI), and Hotwell Level (CM) and Hotwell Level (HMI). And the face plate variable can be linked to COP Min FCV(CM) and COP Min FCV(HMI).

Accordingly, the simulator 100 outputs information about analog monitoring variables and face plate variables through the screen 170 of the operation control panel 110.

The simulator 100 registers tags in the OPC input/output model 150 and the input/output model 160 to compile and generate an executable file.

Figure 6 is a diagram referenced to explain scenario configuration and learning model creation of a plant diagnosis system according to an embodiment of the present invention.

As shown in FIG. 6, the simulator 100 verifies the fidelity of the simulator 100 after a linkage test according to model integration.

The instructor operation panel 120 maintains thermal balance at a rated 100% load for the simulator 100, checks whether key data satisfies the design data in the thermal balance state, and controls the simulator 100 based on the results. Judge fidelity.

The instructor operation panel 120 lowers the load to the thermal balance state at 100% load, switches to the cold state, and operates up to the rated load to check the fidelity of the simulator 100. The model error rate calculation formula is as follows:

Model fidelity is calculated based on the difference between design data and simulation data.

As shown in (a) of FIG. 6, the simulator 100 generates a scenario for each driving condition when the fidelity of the simulator 100 satisfies the set conditions.

The simulator 100 generates scenarios for each operating condition, such as cold state, hot state, gas turbine preheating, startup, heat recovery boiler preheating, steam turbine speed up, grid feed-in, output 50%, output 75%, and output 100%. .

As shown in (b) of FIG. 6, the simulator 100 performs simulation by applying an artificial intelligence-based abnormality diagnosis prediction algorithm that predicts abnormalities and diagnoses failures based on the generated scenario.

The simulator 100 acquires simulator data and stores it cumulatively in the database (DB) 50. The simulator 100 distinguishes between normal and abnormal states and stores each simulation data.

The simulator 100 can perform simulations for each of a plurality of scenarios.

The diagnosis prediction device 200 learns data generated through simulation by applying an abnormality diagnosis prediction algorithm and constructs a learning model based on the learning results.

The diagnostic prediction device 200 analyzes the data of the plant 10 based on the generated learning model, generates diagnostic results for normal and abnormal states, and creates a solution for the diagnostic results.

Accordingly, the artificial intelligence-based plant diagnosis system and its operation method using the simulator of the present invention can generate scenarios for various operating conditions and obtain reliable data on normal and abnormal states through simulation, even if the plant is not actually operated. You can.

In addition, the present invention can predict plant abnormalities for operating conditions that deviate from normal conditions and abnormal conditions.

The present invention has been described with reference to the embodiments shown in the drawings, but these are merely illustrative, and those skilled in the art will recognize that various modifications and other equivalent embodiments can be made therefrom. You will understand. Therefore, the true technical protection scope of the present invention should be determined by the scope of the patent claims below.

20: Steady state simulation data 30: Abnormal state simulation data
50: Database (DB)
100: Simulator 110: Driving control panel
120: Instructor operation panel 130: Control model
140: Process model 150, 160: Input/output model
200: Diagnosis prediction device

Claims (18)

A simulator that configures a process model and control model corresponding to the plant and performs simulation according to operating conditions to generate simulation data; and
A diagnostic prediction device that generates a learning model for the plant based on the simulation data, analyzes the data of the plant based on the learning model, diagnoses the state of the plant, and predicts abnormalities; Including,
The simulator is an artificial intelligence-based plant diagnosis system using a simulator, characterized in that it creates an integrated linkage model by linking input and output variables between the process model and the control model, and creates a scenario for each operating condition to perform simulation. According to claim 1,
The simulator is,
A linkage model including external environmental conditions for the process model and the control model;
An operation control panel that outputs progress according to the simulation; and
an instructor operation panel that runs the simulator to perform simulation; An artificial intelligence-based plant diagnosis system using a simulator that further includes. According to claim 2,
The simulator is an artificial intelligence-based plant diagnosis system using a simulator, characterized in that it tunes the parameters of the process model, modifies the logic of the control model in response to the tuned parameters, and links input and output variables. According to claim 3,
The simulator integrates and links the analog monitoring variables and face plate variables of the control model with the variables of the input/output model, registers a tag in the input/output model, and outputs it on the screen of the operation control panel. Intelligence-based plant diagnostic system. According to claim 1,
The simulator is an artificial intelligence-based plant diagnosis system using a simulator, characterized in that the integrated linkage model is generated by reflecting simulation operation procedure data based on the operation procedure of the plant. According to claim 1,
The simulator is an artificial intelligence-based plant diagnosis system using a simulator, characterized in that the simulator verifies the integrated linkage model by calculating the fidelity of the integrated linkage model by reflecting the design data and operation data of the plant. According to claim 6,
The simulator maintains thermal balance at 100% rated load, and determines fidelity to the integrated linkage model by checking whether key data satisfies the design data in the thermal balance state. It is an artificial intelligence-based system using a simulator. Plant diagnostic system. According to claim 6,
The simulator switches to a cold state by lowering the load to the thermal balance state at 100% load, and starts up to the rated load to determine fidelity to the integrated linkage model. An artificial intelligence-based plant diagnosis system using a simulator. . According to claim 1,
The simulator generates a scenario for each operating condition for at least one of cold state, hot state, gas turbine preheating, startup, heat recovery boiler preheating, steam turbine speed up, grid feed-in, output 50%, output 75%, and output 100%. An artificial intelligence-based plant diagnosis system using a simulator. According to claim 1,
The diagnosis prediction device generates a normal state diagnosis prediction model and an abnormal state diagnosis prediction model based on a learning model generated by learning simulation data generated from the simulator,
Diagnosing and predicting the state of the plant according to operating conditions outside of the normal state based on the normal state diagnosis and prediction model, and diagnosing and predicting failure of the plant for abnormal states based on the abnormal state diagnosis and prediction model. Artificial intelligence-based plant diagnosis system using a featured simulator. According to claim 10,
The diagnostic prediction device diagnoses and predicts the state of the plant according to operating conditions that deviate from the normal state based on the normal state diagnostic prediction model, and detects a failure of the plant due to an abnormal state based on the abnormal state diagnostic prediction model. An artificial intelligence-based plant diagnosis system using a simulator that is characterized by diagnosis and prediction. A simulator configuring a process model and control model corresponding to the plant;
Generating an integrated linkage model by linking input and output variables between the process model and the control model;
Generating simulation data by creating a scenario for each driving condition and performing a simulation;
A diagnostic prediction device generating a learning model for the plant based on the simulation data; and
Diagnosing the state of the plant and predicting abnormalities by analyzing data from the plant based on the learning model; Method of operating an artificial intelligence-based plant diagnosis system using a simulator including. According to claim 12,
The step of creating the integrated linkage model is,
Constructing a linkage model including external environmental conditions for the process model and the control model; and
Integrating the process model, the control model, and the linkage model;
generating the integrated linkage model by reflecting simulation operation procedure data based on the operation procedure of the plant; Method of operating an artificial intelligence-based plant diagnosis system using a simulator including. According to claim 12,
The step of creating the integrated linkage model is,
Tuning parameters of the process model;
Linking input and output variables by modifying the logic of the control model in response to the tuned parameters;
Integrating and linking analog monitoring variables and face plate variables of the control model with variables of the input/output model; and
Registering a tag in the input/output model and outputting it on the operation panel screen of the simulator; Method of operating an artificial intelligence-based plant diagnosis system using a simulator including. According to claim 12,
After the step of creating the integrated linkage model,
Verifying the integrated linkage model by calculating fidelity to the integrated linkage model by reflecting the design data and operation data of the plant; Method of operating an artificial intelligence-based plant diagnosis system using a simulator further comprising: According to claim 15,
The step of verifying the integrated linkage model is,
Setting the simulator to maintain thermal balance at a rated 100% load; and
determining fidelity to the integrated linkage model by checking whether key data obtained in thermal equilibrium satisfies the design data; Method of operating an artificial intelligence-based plant diagnosis system using a simulator including. According to claim 16,
The step of verifying the integrated linkage model is,
The simulator switches to a cold state by lowering the load from the thermal balance state at 100% load; and
After switching, starting up to the rated load and determining fidelity to the integrated linkage model; Method of operating an artificial intelligence-based plant diagnosis system using a simulator further comprising: According to claim 12,
The step of diagnosing the condition of the plant and predicting abnormalities is,
Generating a normal state diagnosis prediction model and an abnormal state diagnosis prediction model based on a learning model generated by learning simulation data generated from the simulator;
Diagnosing and predicting the state of the plant according to operating conditions outside of the normal state based on the steady-state diagnosis and prediction model; and
Diagnosing and predicting a failure of the plant due to an abnormal state based on the abnormal state diagnosis and prediction model; Method of operating an artificial intelligence-based plant diagnosis system using a simulator including.