Deterioration Diagnosis System

Abstract

The purpose is to detect changes in the performance indicators of a facility using equipment within the facility. [Solution] A degradation diagnosis system equipped with a processor, the processor acquires multiple operating data sets which are associated with external conditions determined by external factors of the object to be diagnosed for degradation and values ​​of performance indicators that show the performance of the object under those external conditions, and uses the operating data measured during the period when the object is normal from among the acquired multiple operating data sets as training data to train a learning model in which the external conditions are the explanatory variables and the values ​​of the performance indicators are the dependent variables.

Description

Technical Field

[0001] The present invention relates to a deterioration diagnosis system.

Background Art

[0002] Patent Document 1 discloses a deterioration diagnosis system including: an accumulation unit that accumulates operation data including an operation performance value, which is a performance value obtained by measuring the performance of a device during actual operation, and a condition value indicating environmental conditions during actual operation; an acquisition unit that acquires test data including a pre-deterioration performance value, which is a performance value obtained by testing the performance of the device before deterioration, and a condition value indicating environmental conditions during the test, selects operation data including environmental conditions identical or similar to those during the test from the operation data accumulated in the accumulation unit, and calculates a correction rate from the operation performance value included in the selected operation data and the pre-deterioration performance value; and a diagnosis unit that acquires new operation data and performs deterioration diagnosis of the device based on the operation performance value included in the new operation data and the correction rate.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Facilities such as power plants and factories function through the combination of multiple pieces of equipment to produce electricity, products, and other goods. Such facilities have performance indicators that represent the overall performance of the facility. For example, in the case of a thermal power plant, there is an indicator called thermal efficiency, which shows the amount of electricity generated in relation to the amount of heat from the fuel input into the facility, and indicates the efficiency of the thermal power plant. If the value of this performance indicator decreases, it is desirable to promptly detect and correct any malfunctions in the facility. However, the value of this performance indicator can fluctuate not only due to the influence of equipment inside the facility, but also due to external conditions such as atmospheric temperature and humidity. For example, even if there is no malfunction in the equipment inside the facility, the performance indicator may decrease due to external conditions. In this case, even if an investigation into malfunctions in the equipment inside the facility is conducted in conjunction with the decrease in the performance indicator, the malfunction may not be detected, and investigation costs will be incurred. For example, in large facilities such as thermal power plants, the number of pieces of equipment is large, so investigation costs are high. The present invention aims to detect when the value of a facility's performance indicator changes due to equipment inside the facility. [Means for solving the problem]

[0005] The deterioration diagnosis system to which the present invention is applied includes a processor, the processor which acquires a plurality of operating data sets that associate external conditions determined by external factors of the object to be diagnosed for deterioration with the value of a performance index that indicates the performance of the object under those external conditions, and the system which uses the operating data measured during a period when the object is normal from among the acquired plurality of operating data sets as training data to train a learning model in which the external conditions are the explanatory variables and the value of the performance index is the objective variable. Here, the subject may be defined as a facility capable of measuring performance indicator values. Here, the subject may be a power generation facility, and the performance indicator may be the thermal efficiency of the power generation facility. Here, the operating data measured during the normal period of the subject may be defined as the operating data from the period in which the subject is operating without malfunction.

[0006] From another perspective, the deterioration diagnosis system to which the present invention is applied is a deterioration diagnosis system comprising a processor, wherein the processor uses operating data measured during a normal period of the target as learning data, from among a plurality of operating data sets, in which external conditions determined by external factors outside the target to be diagnosed for deterioration and the value of a performance index indicating the performance of the target under said external conditions are associated, and acquires a learned model with said external conditions as the explanatory variable and said performance index value as the objective variable, and uses the acquired learned model to estimate the value of the performance index from the external conditions of the diagnostic data, which is the operating data used for diagnosis, and outputs information on whether or not the internal equipment of the target is deteriorating based on the estimated value of the performance index and the value of the performance index of the diagnostic data. In this case, if the difference between the estimated performance index value and the performance index value of the diagnostic data is greater than a predetermined value, information indicating that the target has deteriorated may be output. Here, the value of the performance indicator may be estimated using only the aforementioned external conditions.

[0007] Furthermore, from another perspective, the deterioration diagnosis system to which the present invention is applied is a deterioration diagnosis system comprising a processor, wherein the processor acquires a plurality of operating data sets that associate external conditions determined by external factors of the object to be diagnosed for deterioration with the value of a performance index indicating the performance of the object under those external conditions, accepts the specification of a period to be used as learning data for learning from the plurality of operating data sets, and a specification of a period to be used as diagnostic data for diagnosis from the plurality of operating data sets, uses the specified learning data to train a learning model in which the external conditions are explanatory variables and the value of the performance index is the objective variable, uses the trained learning model to estimate the value of the performance index from the external conditions of the specified diagnostic data, and outputs information on whether or not the object is deteriorating based on the estimated value of the performance index and the value of the performance index of the diagnostic data. [Effects of the Invention]

[0008] Equipment within the facility can detect changes in the facility's performance indicators. [Brief explanation of the drawing]

[0009] [Figure 1] This is a block diagram illustrating an example of the deterioration diagnosis system of this embodiment. [Figure 2] This figure shows an example of the hardware configuration for a data storage device, a learning device, and an inference device. [Figure 3] Figure 3(a) shows the functional configuration of the learning device, and Figure 3(b) shows the functional configuration of the inference device. [Figure 4] This diagram illustrates the determination of the degree of deterioration by the judgment unit during normal operation. [Figure 5] This diagram illustrates the determination of the degree of deterioration by the judgment unit when an abnormality occurs. [Figure 6] This figure shows an example of estimated thermal efficiency and measured thermal efficiency. [Figure 7] A flowchart illustrating an example of the learning process flow by a learning device. [Figure 8] This flowchart shows an example of the inference process flow by the inference device. [Figure 9] This flowchart shows an example of the process flow for determining degradation using an inference device. [Modes for carrying out the invention]

[0010] Embodiments of the present invention will be described in detail below with reference to the attached drawings. Figure 1 is a block diagram showing an example of the deterioration diagnosis system 1 of this embodiment. The degradation diagnosis system 1 of this embodiment is a system for diagnosing the degradation of a thermal power plant 3. The degradation diagnosis system 1 comprises a sensor unit 20, a data storage device 30, a learning device 100, and an inference device 200. The learning device 100 has a learning model 160 (see Figure 3(a)). The learning device 100 performs machine learning on the learning model 160 to estimate the thermal efficiency, which is an example of a performance indicator of the thermal power plant 3, using measured values ​​of the external environment during the operation of the thermal power plant 3. The inference device 200 estimates the thermal efficiency of the thermal power plant 3 using the learning model 160 learned by the learning device 100 and diagnoses whether or not degradation has occurred. The learning model 160 learned by the learning device 100 may be referred to as the trained learning model 160. In this embodiment, thermal efficiency is used as a performance indicator, but the performance indicator is not limited to thermal efficiency. Other performance indicators for the thermal power plant 3 could also be, for example, fuel utilization rate or power generation per unit time.

[0011] In the degradation diagnosis system 1, the target for degradation diagnosis is the thermal power plant 3. The thermal power plant 3 generates electricity by burning combustible gas to rotate a turbine and generate steam, which in turn rotates the turbine. The thermal power plant 3 supplies electricity at its rated frequency. Here, the rated frequency is a predetermined value for the frequency of electricity, for example, 50Hz or 60Hz. In this embodiment, the thermal power plant 3 receives natural gas as the combustible gas from the fuel supply equipment 2. The fuel used is not particularly limited, and fuels appropriate to the type of turbine can be used. Examples of turbine types include gas turbines that use hydrogen as fuel, gas turbines that use ammonia as fuel, or gas turbines that use oil as fuel, and fuels appropriate to each can be used.

[0012] Fuel supply facility 2 is, for example, an LNG (liquefied natural gas) receiving terminal. Fuel supply facility 2 is equipped with tanks for storing LNG and pumps for transferring LNG.

[0013] The thermal power plant 3 includes a gas turbine 301, a gas turbine generator 302, a waste heat recovery boiler 303, a steam turbine 304, a steam turbine generator 305, a condenser 306, and an air compression unit 307. The thermal power plant 3 may also include a plurality of other facilities and devices, and a part of these configurations may be integrally formed. More specifically, for example, the gas turbine 301 and the steam turbine 304 may be grouped together to rotate the rotating shaft of one generator to generate electricity.

[0014] The gas turbine 301 mixes a combustible gas such as natural gas supplied from the fuel supply facility 2 with air and burns them, and rotates the turbine with the high-temperature and high-pressure combustion gas. The gas turbine generator 302 generates electricity using the rotational force of the gas turbine 301. The waste heat recovery boiler 303 recovers the waste heat contained in the exhaust gas of the gas turbine 301 and generates steam. The steam turbine 304 rotates the turbine using the steam generated by the waste heat recovery boiler 303. The steam turbine generator 305 generates electricity using the rotational force of the steam turbine 304. The condenser 306 is arranged at the outlet of the steam turbine 304 and cools the steam after rotating the steam turbine 304 and returns it to water. The condenser 306 includes a heat exchanger, and a cooling medium is circulated through the heat exchanger to cool the steam. Note that the cooling medium is not particularly limited, and for example, liquids such as seawater, river water, industrial water, or gases such as air can be used. The air compression unit 307 is a device that compresses the air in the atmosphere and supplies it to the gas turbine 301.

[0015] The sensor unit 20 includes a plurality of measuring instruments for collecting data on the external environment of the thermal power plant 3. Here, the external environment refers to natural conditions and operating conditions that are not controlled by the internal devices of the thermal power plant 3. Natural conditions include, for example, atmospheric temperature, atmospheric humidity, atmospheric pressure, and the temperature of seawater and river water. Operating conditions are artificially determined and are not related to the deterioration of the equipment of thermal power plant 3. Operating conditions include, for example, the rated frequency required for the power plant's electricity. Other operating conditions include, for example, the fuel gas flow rate and the heat content of the fuel gas supplied to thermal power plant 3.

[0016] The sensor unit 20 includes a barometer 21, an air density meter 22, an air thermometer 23, an air hygrometer 24, and a water thermometer 25 as measuring devices for measuring the external environment. Furthermore, the sensor unit 20 includes a gas flow meter 26, a gas calorimeter 27, an electrical frequency meter 28, and an energy meter 29 as measuring devices for measuring operating conditions.

[0017] The barometer 21 measures the atmospheric pressure around the thermal power plant 3. The air density meter 22 measures the air density around the intake port of the air supplied to the air compression unit 307. The air thermometer 23 measures the temperature of the air around the intake port of the air supplied to the air compression unit 307. The air hygrometer 24 measures the humidity of the air surrounding the thermal power plant 3. The thermometer 25 measures the temperature of the seawater or river water supplied to the condenser 306. The gas flow meter 26 measures the flow rate of combustible gases, such as natural gas, supplied to the thermal power plant 3. The gas flow meter 26 is installed, for example, in a pipe connecting the fuel supply equipment 2 and the thermal power plant 3. The gas calorimeter 27 measures the amount of heat per unit of combustible gases such as natural gas. The gas calorimeter 27 is installed, for example, in the middle of a pipe connecting the fuel supply facility 2 to the thermal power plant 3. The electrical frequency meter 28 measures the frequency of the electricity supplied by the thermal power plant 3. The electricity meter 29 measures the amount of electricity supplied by the thermal power plant 3.

[0018] These multiple measuring instruments take measurements at predetermined intervals, for example, every second or every minute. These intervals may be the same for all the measuring instruments, or they may be different.

[0019] The data storage device 30 acquires the measured values ​​measured by each measuring instrument of the sensor unit 20 and stores them in the information storage device 202, which will be described later. The information stored in the data storage device 30 may be raw sensor data or processed raw data. Processed data, for example, if minute-by-minute driving data is used as learning data as described later, may be the average value or integral value over one minute of raw data measured every second.

[0020] Figure 2 shows an example of the hardware configuration of the data storage device 30, the learning device 100, and the inference device 200. As shown in Figure 2, the computer that implements the data storage device 30, the learning device 100, and the inference device 200 has a processing unit 201 that performs digital arithmetic processing according to a program. The computer that implements the data storage device 30, the learning device 100, and the inference device 200 also has an information storage device 202 for storing information and a network interface 203 for enabling communication via a LAN (=Local Area Network) cable or the like.

[0021] The processing unit 201 is comprised of a computer. The processing unit 201 includes a CPU (=Central Processing Unit) 211, which is an example of a processor that performs various processes. The processing unit 201 also includes a ROM (Read Only Memory) 212 where programs are stored, and a RAM (Random Access Memory) 213 used as a work area. The information storage device 202 is implemented using existing devices such as a hard disk drive, semiconductor memory, or magnetic tape. The processing unit 201, the information storage device 202, and the network interface 203 are connected via a bus 206 and signal lines (not shown).

[0022] The program executed by the CPU 211 can be provided to the learning device 100 while stored on a computer-readable recording medium such as a magnetic recording medium (magnetic tape, magnetic disk, etc.), an optical recording medium (optical disk, etc.), a magneto-optical recording medium, or semiconductor memory. Alternatively, the program executed by the CPU 211 may be provided to the learning device 100 using communication means such as the Internet.

[0023] In this embodiment, each process is executed on any computer. This computer may be implemented as a processor (hardware), a program (software), or a combination thereof. The computer may be a general-purpose computer, a computer designed for a specific purpose, a workstation, or any other system capable of executing each process.

[0024] The processor is configured to perform various processes in cooperation with the program. The processor can function as each unit or each means in this embodiment. The execution order of the processes performed by the processor is not limited to the order described in this embodiment and can be changed as needed.

[0025] A processor can be configured with one or more hardware components. The types of hardware that make up a processor are not limited to any particular type. For example, a processor may be a CPU211, an MPU (=Micro Processing Unit), a programmable logic device such as an FPGA (=Field Programmable Gate Array), a dedicated circuit for performing specific processing such as an ASIC (=Application Specific Integrated Circuit), a GPU (=Graphic Processing Unit), or hardware such as an NPU (=Neural Processing Unit).

[0026] A processor can be configured not only with a combination of multiple hardware of the same type, but also with a combination of multiple hardware of different types. When multiple hardware is configured to execute one or more processes of a processor, the multiple hardware may reside in physically separate devices or in the same device. Hardware is composed of electrical circuits, etc., which are combinations of circuit elements such as semiconductor elements. In any embodiment, the execution order of each process by the processor is not limited to the order described in each embodiment, and can be changed as necessary.

[0027] The program may be firmware or software such as microcode. The program may also be, for example, a group of program modules. Each function constituting the group of program modules may be implemented by a processor configured to execute each function. In each embodiment, the program may be program code or multiple code segments stored in one or more non-temporary computer-readable media (e.g., semiconductor memory, magnetic or optical storage media, or other storage).

[0028] A program may be divided and stored on multiple non-temporary computer-readable media located on devices that are physically separated from each other. Program code and multiple code segments may be represented by any combination of procedures, functions, subprograms, routines, subroutines, modules, software packages, classes, instructions, data structures, and program statements. Program code and multiple code segments may be connected to other code segments or hardware circuits by sending and receiving information, data, arguments, parameters, or memory contents.

[0029] Figures 3(a) and 3(b) show the functional configuration of the learning device 100 or the inference device 200. Figure 3(a) shows the functional configuration of the learning device 100. Figure 3(b) shows the functional configuration of the inference device 200. The learning device 100 is a device for generating parameters for an inference model that estimates thermal efficiency, using learning data in which operating data and thermal efficiency are associated. The learning device 100 has a reception unit 110 that receives input from the user. The learning device 100 also has an acquisition unit 120 that acquires operation data from the data storage device 30. The learning device 100 also has a sorting unit 130 that sorts the operation data into normal data and abnormal data. Furthermore, the learning device 100 has a preprocessing unit 140 that processes the operation data into a format that can be input into the learning model 160 and generates learning data. Furthermore, the learning device 100 has an input unit 150 that inputs the learning data into the learning model 160. The learning device 100 also has a learning model 160 that is implemented using a neural network or the like. Furthermore, the learning device 100 has a setting unit 170 that sets hyperparameters.

[0030] The reception unit 110 receives various information from the user. The reception unit 110 also receives a specification from the user regarding the period of operating data to be used as learning data. The timing of the learning process by the learning device 100 is not particularly limited. That is, the learning process may be executed at a timing arbitrarily specified by the user, or it may be executed automatically by the device. In this case, the acceptance of the period of operating data to be used as learning data may be specified by the user, or a predetermined period as the period of operating data may be accepted. For example, if daily learning is performed, it may be specified that operating data up to one year prior to the day the learning of operating data is performed be used. Furthermore, the period of operating data may be automatically determined according to predetermined conditions. The acquisition unit 120 acquires operating data from the data storage device 30. Furthermore, the sorting unit 130 sorts the operating data into normal data and abnormal data. Here, normal data refers to operating data taken when the thermal power plant 3 is operating normally. Whether or not the thermal power plant 3 is operating normally is determined by the manager of the thermal power plant 3. The manager determines whether or not it is operating normally, and the period that the manager deems to be normal can be considered the period of normal operation. In addition, the sorting unit 130 may sort the operating data into normal data and abnormal data based on predetermined conditions. Here, "predetermined conditions" can refer, for example, to a period in which the frequency fluctuation amount of the generator is smaller than a predetermined threshold. That is, operating data measured when the thermal power plant 3 is operating at its rated frequency is sorted as normal data.

[0031] The preprocessing unit 140 formats, transforms, and organizes the operating data that the sorting unit 130 has sorted as normal data, into a format suitable for learning data to be input to the learning model 160. The preprocessing unit 140 treats the measured values ​​as one learning data unit using the time unit used for the learning data to be input to the learning model 160, and assigns the thermal efficiency, which is the target variable, to each learning data unit. Here, the time unit may be, for example, one second or one minute.

[0032] In this embodiment, examples of external conditions used as explanatory variables for training data include atmospheric pressure, air density at the air compressor inlet, air temperature at the air compressor inlet, atmospheric humidity, and circulating water temperature. The external conditions used as explanatory variables are not limited to these, and other external conditions may be used.

[0033] Thermal efficiency is information that indicates the efficiency of the thermal power plant 3. Thermal efficiency is the ratio that shows how much of the thermal energy of the fuel input is extracted as electrical energy. Thermal efficiency (%) is expressed by the formula: Thermal efficiency (%) = (Electrical energy obtained by power generation ÷ Thermal energy of the fuel input) × 100. More specifically, the value obtained by multiplying the measurement value of the gas flow meter 26 (see Figure 1), which measures the flow rate of combustible gas such as natural gas supplied to the thermal power plant 3, by the measurement value of the gas calorimeter 27 (see Figure 1), which represents the amount of heat per unit of the supplied gas, is the "thermal energy of the fuel input". In addition, the amount of electricity measured by the electricity meter 29 can be used to obtain the "electrical energy obtained by power generation".

[0034] Furthermore, the preprocessing unit 140 performs tasks such as filling in missing values ​​and removing abnormal or outlier values. The input unit 150 inputs the preprocessed training data from the preprocessing unit 140 into the training model.

[0035] The learning model 160 is trained to infer thermal efficiency based on external conditions included in the operating data. In this embodiment, the learning model 160 utilizes a regression model. Various learning models such as linear regression, ridge regression, lasso regression, decision tree regression, and SVR can be used as regression models.

[0036] The hyperparameters set by the setting unit 170 include, for example, the number of training iterations, the learning rate, and the batch size. The setting unit 170 has a function to receive instructions from the user to adjust the hyperparameters. The setting unit 170 may also be equipped with an algorithm that automatically adjusts some or all of the hyperparameters, thereby automatically adjusting the hyperparameters.

[0037] Furthermore, as shown in Figure 3(b), the inference device 200 has a reception unit 210 that receives input from the user. The inference device 200 has an inference target data acquisition unit 220 that acquires the data to be inferred from the data storage device 30. The inference device 200 also has a trained learning model acquisition unit 230 that acquires the trained learning model 160 from the learning device 100. Furthermore, the inference device 200 has a pre-processing unit 240 that processes the data to be inferred so that it can be input into the learning model. The inference device 200 also has an input unit 250 that inputs the processed data to be inferred into the trained learning model 160. The inference device 200 has a determination unit 260 that determines the degree of deterioration of the thermal power plant 3 based on the thermal efficiency output from the trained learning model 160. The inference device 200 has a notification unit 270 that notifies the user based on the determination of the determination unit 260.

[0038] The preprocessing unit 240 performs the same processing on the data to be inferred as the preprocessing unit 140 of the learning device 100.

[0039] When the input unit 250 receives the data to be inferred, which has been processed by the preprocessing unit 240, for the trained model 160, the trained model 160 outputs the thermal efficiency estimated based on the data to be inferred.

[0040] The determination unit 260 determines whether the thermal power plant 3 has deteriorated based on the thermal efficiency estimated by the trained learning model 160 and the thermal efficiency based on the measured value. The criteria used to determine deterioration are not particularly limited. For example, if the difference between the estimated performance index value and the performance index value from the diagnostic data is greater than a predetermined value, it may be determined that the thermal power plant 3 has deteriorated. For example, the estimated thermal efficiency value and the measured thermal efficiency value can be used to determine deterioration using a value calculated by a predetermined formula. Specifically, the determination unit 260 may determine that the thermal power plant 3 has deteriorated if the value obtained by subtracting the measured thermal efficiency value from the estimated thermal efficiency value is greater than a predetermined threshold. Alternatively, the determination unit 260 may determine that the thermal power plant 3 has deteriorated if the value obtained by dividing the estimated thermal efficiency value by the measured thermal efficiency value is less than a predetermined threshold. In other words, the deterioration of the thermal power plant 3 may be determined according to the ratio between the estimated performance index value and the performance index value from the diagnostic data. Furthermore, the determination unit 260 may output information indicating not only whether the thermal power plant 3 has deteriorated, but also the degree of deterioration. Specifically, the degree of deterioration may be pre-associated with the magnitude of the value obtained by subtracting the measured thermal efficiency value from the estimated thermal efficiency value, and the determination unit 260 may output the degree of deterioration as information indicating the degree of deterioration.

[0041] The notification unit 270 displays information regarding the deterioration determined by the determination unit 260, for example, on the user interface used by the user. When the determination unit 260 determines that the thermal power plant 3 is deteriorating, the notification unit 270 may also notify the user to investigate whether there are any malfunctions in the internal equipment of the thermal power plant 3. The means of notifying the user are not limited to this, and for example, voice notifications using a speaker may also be provided.

[0042] Figure 4 is a diagram illustrating the determination of the degree of deterioration by the determination unit 260 during normal operation. As described above, the determination unit 260 determines the deterioration of the thermal power plant 3 by comparing the measured thermal efficiency with the thermal efficiency estimated using external conditions. Figure 4 shows the measured thermal efficiency 901 and the estimated thermal efficiency 902 during normal operation.

[0043] The measured thermal efficiency of 901 is calculated based on the measured amount of thermal energy supplied to the thermal power plant 3 and the measured amount of power generated by the thermal power plant 3. Specifically, as described above, it is calculated by dividing the measured amount of power generated, which is the electrical energy obtained by power generation, by the measured amount of thermal energy, which is the thermal energy of the fuel input, as the numerator.

[0044] On the other hand, the estimated thermal efficiency is calculated by inputting external variables such as air temperature and temperature into a pre-trained model 160, resulting in an estimated thermal efficiency of 902. As mentioned above, the trained model 160 is trained based on normal operating data, so the estimated thermal efficiency 902 will be an estimate of the thermal efficiency value when the thermal power plant 3 is operating normally. Therefore, if thermal power plant 3 is functioning normally, there is a high probability that the measured thermal efficiency 901 and the estimated thermal efficiency 902 will be the same or similar approximate values. This probability can be quantified through statistical processing and may be expressed, for example, as a confidence level or confidence interval.

[0045] Figure 5 is a diagram illustrating the determination of the degree of deterioration by the determination unit 260 when an abnormality occurs. Figure 5 shows the state of thermal power plant 3 when an abnormality occurs inside, causing a change in its internal structure. When an abnormality occurs, the amount of electricity generated decreases. As a result, the measured value of the generated energy decreases, and the measured thermal efficiency 901 decreases. However, if, for example, a limiter that suppresses the output malfunctions, the amount of electricity generated may increase as a result of the abnormality. On the other hand, with the estimated thermal efficiency 902, the internal state of the thermal power plant 3 is not input into the trained model 160. Therefore, even if an abnormality occurs in the thermal power plant 3, the same estimation as under normal conditions is performed. As a result, there is a difference between the measured thermal efficiency 901 and the estimated thermal efficiency 902. Therefore, if there is a difference between the measured thermal efficiency 901 and the estimated thermal efficiency 902, it is determined that the thermal power plant 3 is deteriorating.

[0046] Figure 6 shows an example of the estimated thermal efficiency and the measured thermal efficiency. In Figure 6, the measured values ​​of thermal efficiency are shown on the horizontal axis, and the estimated values ​​of thermal efficiency are shown on the vertical axis. In Figure 6, a straight line R with a slope of 1 is drawn passing through the origin, and at points on this line R, the measured values ​​of thermal efficiency and the estimated values ​​of thermal efficiency coincide. In Figure 6, the relationship between the measured and estimated thermal efficiency when the training data is used is shown by black circles. The relationship between the measured and estimated thermal efficiency when the test normal data is used is shown by a dashed white line. The relationship between the measured and estimated thermal efficiency when the test abnormal data is used is shown by a solid white line.

[0047] Here, test data refers to data used to verify whether the trained model 160 has been trained correctly, and in this case, actual operating data acquired in the past is used. Test normal data refers to operating data measured when thermal power plant 3 was operating normally from past operating data. Test abnormal data refers to operating data measured when thermal power plant 3 was operating abnormally from past operating data.

[0048] As shown in the figure, the normal training data and normal test data are positioned close to the line R, while the abnormal test data are positioned far from the line R. This confirms that when an anomaly occurs, the measured value and the estimated value will differ.

[0049] Figure 7 is a flowchart showing the flow of the learning process by the learning device 100. First, the acquisition unit 120 of the learning device 100 reads past operating data from the data storage device 30 (step 401). Next, the sorting unit 130 extracts normal data from the acquired operating data (step 402). Next, the reception unit 110 receives the user's specification of the learning data (step 403). After that, the preprocessing unit 140 performs preprocessing on the learning data for the period specified by the user and creates input variables (step 404).

[0050] Then, the reception unit 110 determines whether or not to perform hyperparameter search from the user (step 405). If hyperparameter search is to be performed (YES in step 405), the process proceeds to step 406, where the setting unit 170 performs hyperparameter search and saves them. Then, the learning model 160 reads the saved hyperparameters (step 407). Then, the learning model 160 performs machine learning using the training data preprocessed by the preprocessing unit 140, adjusts the model parameters of the trained machine learning model that estimates thermal efficiency, saves the adjusted model parameters (step 408), and the learning process ends.

[0051] In step 405, if it is decided not to search for new hyperparameters (NO in step 405), the process proceeds to step 407 without searching for hyperparameters, and the saved hyperparameters are loaded (step 407). Then, the learning model 160 performs machine learning using the training data preprocessed by the preprocessing unit 140, adjusts the model parameters of the trained machine learning model that estimates thermal efficiency, saves the adjusted model parameters (step 408), and the learning process ends.

[0052] Figure 8 is a flowchart showing the flow of the inference process by the inference device 200. First, the data acquisition unit 220 of the inference device 200 reads the driving data to be inferred for a period specified by the user from the data storage device 30 (step 501). Then, the preprocessing unit 240 preprocesses the driving data, creates input variables, and converts the driving data into a format that can be input into the trained model 160 (step 502). After that, the trained model acquisition unit 230 reads the saved model parameters of the trained model 160 (step 503). At this time, the trained model acquisition unit 230 may also read hyperparameter information.

[0053] The input unit 250 then inputs the operating data processed by the preprocessing unit 240 as data to be inferred into the trained model 160, and estimates the thermal efficiency of the data to be inferred (step 504). The estimated thermal efficiency is then output to the determination unit 260 (step 505), and the inference process ends.

[0054] Figure 9 shows an example of a flowchart for determining deterioration. The determination unit 260 obtains the estimated value of thermal efficiency estimated by the inference device 200 and the measured value of thermal efficiency from the operating data of the data to be inferred (step 601). The determination unit 260 takes the difference between the measured value and the estimated value (step 602). The determination unit 260 determines the degree of deterioration as the difference between the measured value and the estimated value (step 603). The notification unit 270 notifies the user of the degree of deterioration (step 604).

[0055] In this embodiment, external conditions are measured while the thermal power plant 3 is in operation to acquire operating data, and the thermal efficiency, which serves as a performance indicator, is calculated by actual measurement. Then, the operating data measured during periods when the thermal power plant 3 is operating normally is used as training data, and a learning model is trained with external conditions as explanatory variables and the performance indicator value as the dependent variable. Furthermore, by using this pre-trained model to estimate thermal efficiency under external conditions, it becomes possible to estimate fluctuations in thermal efficiency in response to external conditions, thereby reducing unnecessary equipment inspections.

[0056] Furthermore, in this embodiment, the user can set the learning period and the inference period. For example, the thermal power plant 3 has many pieces of equipment, and some kind of change is frequently made, such as repairing or replacing one of them. Therefore, even if the learning model 160 is trained using learning data from a specific normal period, if a significant change in the equipment configuration occurs afterward, the equipment configuration on which the learning model 160 was based at the time of learning may differ from the equipment configuration at the time of actual inference. In such cases, the estimation accuracy may decrease. In this embodiment, by specifying the learning period, the decrease in estimation accuracy can be suppressed by performing learning with operating data after a significant equipment change. In addition, by training with new data at short intervals, such as daily or weekly, it becomes possible to detect short-term abnormal signs. Furthermore, by using past data (for example, data from several years ago) for training, it can be utilized for analyzing long-term aging deterioration.

[0057] Furthermore, the learning device 100 according to this embodiment enables black-box-like diagnosis by using only externally observed environmental conditions and not internal equipment state variables. In particular, in cases where there are many pieces of equipment, such as in large plants, and it is difficult to obtain detailed information about the internal state of the equipment, the learning device 100 according to this embodiment is effective from the standpoint of variable reduction and ease of implementation.

[0058] In this embodiment, a thermal power plant 3 was used as an example to explain the target for deterioration diagnosis, but the target for deterioration diagnosis is not limited to this. Any facility on which performance indicators can be measured can be used as the target for deterioration diagnosis. For example, power plants such as nuclear power plants and hydroelectric power plants, as well as factories and processing facilities, can be given as examples. In addition, devices such as generators, boilers, heat sources, and storage batteries can be given as examples of targets for deterioration diagnosis. Furthermore, the performance indicators are appropriately selected according to the target for deterioration diagnosis. [Explanation of Symbols]

[0059] 1…Degradation diagnosis system, 2…Fuel supply equipment, 3…Thermal power plant, 20…Sensor unit, 30…Data storage device, 100…Learning device, 160…Learning model, 200…Inference device, 220…Inference target data acquisition unit, 230…Learning model acquisition unit, 260…Determination unit, 270…Notification unit

Claims

1. The processor comprises, In a facility where multiple devices are connected and deterioration is to be diagnosed, multiple operational data sets are acquired that associate external conditions determined by external factors of the facility with performance indicator values ​​that show the overall performance of the facility under those external conditions. Using the operational data measured during the period when the facility was operating normally, from among the multiple operational data acquired, A degradation diagnosis system that trains a learning model with the aforementioned external conditions as explanatory variables and the value of the aforementioned performance indicator as the objective variable.

2. The facility is capable of measuring the values ​​of the performance indicators. The deterioration diagnosis system according to claim 1.

3. The deterioration diagnosis system according to claim 1, wherein the facility is a power generation facility, and the performance index is the thermal efficiency of the power generation facility.

4. The operating data measured during the normal period of the aforementioned facility refers to the operating data from the period when the facility was operating without any malfunctions. A deterioration diagnosis system according to any one of claims 1 to 3.

5. Equipped with a processor, The aforementioned processor, In a facility with multiple connected devices, the external conditions determined by external factors outside the facility being diagnosed for deterioration, and the performance index values ​​indicating the overall performance of the facility under those external conditions are correlated. Using the operational data measured during a period when the facility was functioning normally as training data, a trained model is obtained with the external conditions as the explanatory variable and the performance index values ​​as the dependent variable. Using the acquired learning model, the value of the performance index is estimated from the external conditions of the diagnostic data, which is the driving data used for diagnosis. A deterioration diagnostic system that outputs information regarding whether or not the equipment inside the facility is deteriorating, based on the estimated value of the performance index and the value of the performance index in the diagnostic data.

6. If the difference between the estimated performance index value and the performance index value in the diagnostic data is greater than a predetermined value, information indicating that the facility has deteriorated is output. The deterioration diagnosis system according to claim 5.

7. The value of the performance indicator is estimated using only the aforementioned external conditions. The deterioration diagnosis system according to claim 5.

8. Equipped with a processor, The aforementioned processor, In a facility where multiple devices are connected and deterioration is to be diagnosed, multiple operational data sets are acquired that associate external conditions determined by external factors of the facility with performance indicator values ​​that show the overall performance of the facility under those external conditions. The system accepts the specification of a period from the aforementioned multiple operational data to be used as learning data, and a period from the aforementioned multiple operational data to be used as diagnostic data for diagnosis. Using the specified training data, a learning model is trained with the external conditions as explanatory variables and the performance index values ​​as the objective variable; and using the trained learning model, the performance index values ​​are estimated from the external conditions of the specified diagnostic data. A deterioration diagnostic system that outputs information regarding whether or not the facility is deteriorating based on the estimated value of the performance indicator and the value of the performance indicator in the diagnostic data.

Images