JP6116466B2 - Plant diagnostic apparatus and diagnostic method - Google Patents

Description

  The present invention relates to a plant diagnosis apparatus and diagnosis method.

  When an abnormal transient or accident occurs in the plant, the plant diagnosis device detects the occurrence of the abnormality or accident based on the measurement data from the plant.

  Patent Document 1 discloses a diagnostic apparatus that uses adaptive resonance theory (ART). Here, the adaptive resonance theory ART is a theory for classifying multi-dimensional data into categories according to their similarity.

  In the technique of Patent Literature 1, first, normal measurement data is classified into a plurality of categories (normal categories) using adaptive resonance theory ART. Next, the current measurement data is input to the adaptive resonance theory ART and classified into categories. When this measurement data cannot be classified into normal categories, a new category (new category) is generated. The occurrence of a new category means that the state of the plant has changed. Therefore, the occurrence of an abnormality is determined based on the occurrence of a new category, and an abnormality is diagnosed when the occurrence rate of the new category exceeds a threshold value.

  However, the new category is generated when the current measurement data cannot be classified into the normal category, that is, when the state of the plant changes.

  The state of the plant also changes when the operation pattern changes. Here, the operation pattern is not directly related to the equipment characteristics such as the environmental conditions (atmospheric temperature, humidity, etc.) of the place where the plant is installed and the amount determined by the operator's operation (power generation amount generated in the plant). It is an external environmental factor. Regardless of whether or not an abnormality has occurred, the value of various measurement data changes when the operation pattern changes.

  However, although the driving pattern change is not abnormal, there is a possibility that the method of Patent Document 1 diagnoses it as abnormal. A normal state is diagnosed as abnormal, and as a result, false alarms are generated.

  In order to solve this problem, the technique of Patent Document 2 discloses a technique for suppressing the false alarm rate by distinguishing between operation pattern changes and abnormalities. In the present technology, the measurement value for determining the driving pattern is processed by the adaptive resonance theory ART, and it is determined that the driving pattern has changed when a new category is generated.

  Non-Patent Document 1 describes a Bayesian learning method as a method for estimating a representative value of an adjustment parameter and a fluctuation range (standard deviation).

JP 2005-165375 A International publication patent WO2011-125130

Frontier 12 of Statistical Science, Computational Statistics Published October 28, 2005, Yukito Iba and others Iwanami Shoten P222-P232

  In the technique of Patent Document 2, it is determined whether or not the driving pattern has changed, but it is not possible to determine whether or not an abnormality has occurred based on the driving pattern. In addition, long-term operation data is required to learn the normal state of all operation patterns.

  An object of the present invention is to provide a plant diagnosis apparatus and a diagnosis method capable of improving diagnosis accuracy by performing diagnosis corresponding to an operation pattern and diagnosing with short-term operation data. .

  From the above, the diagnostic apparatus according to the present invention is a plant diagnostic apparatus that includes a detection unit that diagnoses the operation state of the plant using a measurement signal obtained by measuring the state quantity of the plant, and displays the diagnosis result on the image display device. The detection unit detects a change in state using a diagnosis model constructed with data for a predetermined period, and an abnormality occurs using a diagnosis model constructed with data of the same operation pattern as when the state change was detected. It is comprised by the abnormality detection part which detects the presence or absence.

  The diagnosis method of the present invention is a plant diagnosis method for diagnosing the operation state of a plant using a measurement signal obtained by measuring the state quantity of the plant and displaying the diagnosis result, and is a diagnosis model constructed with data of a predetermined period Is used to detect the occurrence of an abnormality using a diagnosis model constructed with data of the same operation pattern as when the state change was detected.

  Furthermore, the diagnosis method of the present invention is a plant diagnosis method for diagnosing an operation state of a plant using a measurement signal obtained by measuring a state quantity of the plant and displaying a diagnosis result, and a diagnosis model constructed by data of a predetermined period If a new category occurs, check if there is a change in the plant operation pattern, search for a period of the same operation pattern as the time when the new category was detected, and measure the signal in the operation pattern for that period Is used to construct a diagnostic model, evaluate the presence / absence of a new category in the diagnostic model, and perform abnormality determination.

  Diagnosis accuracy can be improved by performing diagnosis corresponding to the driving pattern. In addition, since the operation data of the diagnosis target plant may be a short period, the time for introducing the diagnosis apparatus can be shortened.

The block diagram explaining the diagnostic apparatus 400 which is 1st Example of this invention. The figure explaining the timing which performs the model construction process mode and diagnostic mode of the diagnostic apparatus. The figure explaining the structure of the data classification function of the classification | category part 750. FIG. The block diagram which shows the structure of F0 layer 621. FIG. The block diagram which shows the structure of F1 layer 622. FIG. The figure explaining the result of having classified a measurement signal or a pseudo signal. The figure which shows an example of the classification result classified into the category. FIG. 6 is an operation flowchart diagram of a diagnosis mode of the diagnosis apparatus 400. The figure explaining the aspect of the data preserve | saved at the measurement signal database 510. FIG. The figure explaining the aspect of the data preserve | saved at the plant database. The figure explaining the aspect of the data preserve | saved at the diagnostic model database. 5 is a block diagram illustrating a pseudo signal generation unit 730. FIG. The figure which shows the example which displayed the example of the parameter data produced | generated by the parameter data 735 on the display screen form. The figure explaining the parameter sample value obtained as a result of operating parameter estimating part 734. The figure which shows the example of the pseudo data produced with the simulator 732. FIG. The figure which displayed two items of pseudo data when changing a parameter. The figure which displayed two items of pseudo data at the time of creating pseudo data using an average value. The figure which shows the screen which displays the determination result in the flowchart of FIG. The figure which shows the screen which displays the measurement signal of a similar plant, a pseudo signal, the result which classified the pseudo signal into the category, and the design information of a diagnosis object plant and a similar plant. The figure which shows the screen used for changing the vigilance parameter (rho) which determines the classification | category performance of a category. The figure explaining the structure in case the plant 100 of a diagnosis object is a gas turbine plant. The figure which shows the variation of the driving | running pattern at the time of starting. The figure explaining the structure in case the plant 100 of a diagnosis object is a boiler plant. The figure which shows the variation of the operation pattern of a boiler plant. The figure explaining the structure in the case of diagnosing a some site.

  A plant diagnosis apparatus and a diagnosis method according to an embodiment of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram for explaining a diagnostic apparatus 400 according to the first embodiment of the present invention. The plant 100 is monitored using the diagnostic device 400.

  The diagnostic device 400 includes a monitoring data generation unit 700, a classification unit 750, and a detection unit 800 as arithmetic devices. The monitoring data generation unit 700 includes an operation pattern evaluation unit 710, a similar plant selection unit 720, and a pseudo signal generation unit 730. The detection unit 800 includes a state change detection unit 810 and an abnormality detection unit 820.

  The diagnostic apparatus 400 includes a measurement signal database 510, a plant database 520, and a diagnostic model database 530 as databases. In FIG. 1, the database is abbreviated as DB. In the measurement signal database 510, the plant database 520, and the diagnostic model database 530, computerized information is stored, and the information is normally stored in a form called an electronic file (electronic data).

  Further, the diagnostic apparatus 400 includes an external input interface 410 and an external output interface 420 as external interfaces.

  And the measurement signal 1 which measured the various state quantities which are the operation states of the plant 100 via the external input interface 410, and the external input device 910 comprised of the keyboard 920 and the mouse 930 provided in the operation management room 900. The external input signal 2 created by the operation is taken into the diagnostic apparatus 400. In addition, the image display information 11 is output to the image display device 940 provided in the operation management room 900 via the external output interface 420.

  In the diagnostic apparatus 400 shown in FIG. 1, the measurement signal 1 obtained by measuring various state quantities of the plant 100 is taken in via the external input interface 410. The measurement signal 3 captured by the diagnostic device 400 is stored in the measurement signal database 510. Although most of the measurement signal 1 is an analog quantity, the measurement signal 3 taken into the diagnostic apparatus 400 is a time-series digital quantity, so that the symbols are distinguished.

  The diagnostic apparatus 400 has two processing modes, model construction processing and diagnostic processing. In the model construction process, a diagnostic model is constructed using the measurement signal in the normal state. The diagnostic model is constructed by clustering technology, and the measurement signals in the normal state are classified into several data groups. In the diagnosis process, a measurement signal at the time of diagnosis is processed. If the measurement signal has the same characteristics as in the normal state, it is classified into one of the data groups of the diagnostic model. If the characteristics are different, they do not belong to the diagnostic model data group.

  When an abnormality occurs in the plant 100, the characteristics of the measurement signal are different from the normal state. For this reason, the diagnostic measurement signal is not classified into the data group of the diagnostic model constructed using the measurement signal in the normal state. Using this characteristic, in Patent Document 1 described above, an abnormality is diagnosed when the measurement signal does not belong to the data group of the diagnostic model.

  However, the characteristics of the plant 100 change according to the operation pattern of the plant. Here, the operation pattern is directly related to the device characteristics such as the environmental conditions (atmospheric temperature, humidity, etc.) of the place where the plant is installed, and the values determined by the operator's operation (command value for the amount of power generated by the plant). It is an unrelated external environmental factor.

  Even if the driving state is normal, the tendency of the measurement signal changes if the driving pattern is different. For this reason, when the technique of Patent Document 1 is applied, there is a possibility that a false report that determines the normal state as abnormal may occur.

  Therefore, in the present invention, the occurrence of false alarms is avoided by diagnosing by combining two types of detection methods, namely, state change detection and abnormality detection.

  First, the operation in the model construction processing mode will be described. In the model construction processing mode, a monitoring model used for detecting a state change is constructed.

  First, the monitoring data generation unit 700 extracts measurement signal information 4 for a predetermined period from the measurement signal database 510 for a predetermined data item. The predetermined period can be arbitrarily set according to the operation cycle of the plant 100. One day for DSS (Daily Start and Stop) operation, one week for WSS (Weekly Start and Stop) operation, MSS (Monthly) In the case of Start and Stop) operation, it is set to one month or the like. The monitoring data information 6 including the measurement signal information 4 extracted by the monitoring data generation unit 700 is transmitted to the classification unit 750. The classification unit 750 classifies the monitoring data information 6 using a clustering technique. The classification result information 7 as the classification result is transmitted to the diagnostic model database 530 and stored.

  Next, the operation in the diagnostic processing mode will be described. In the diagnosis processing mode, first, the state change detection unit 810 determines whether there is a state change using the diagnosis model constructed in the model construction processing mode.

  When a state change is detected, a diagnosis model is constructed using data of the same operation pattern, and the abnormality detection unit 820 determines whether there is an abnormality. If there is no data with the same operation pattern, pseudo data of normal data is created using past operation data of a similar plant, and diagnosis is performed using a diagnostic model constructed using the pseudo data. Details of the diagnostic processing mode will be described later with reference to FIG. Detection result information 10 obtained by operating the diagnostic processing mode is output to the external output interface 420.

  In the diagnostic apparatus 400 of this embodiment, a monitoring data generation unit 700, a classification unit 750, a detection unit 800, a measurement signal database 510, a plant database 520, and a diagnostic model database 530 are provided in the diagnostic apparatus 400. However, some of these devices may be arranged outside the diagnostic device 400 so that only data is communicated between the devices.

  Although not shown, the information stored in the measurement signal database 510, the plant database 520, and the diagnostic model database 530 installed in the diagnostic apparatus 400 can communicate with the operation management room 900, Database information can be displayed on the image display device 940. These pieces of information can be corrected by an external input signal 2 generated by operating an external input device 910 including a keyboard 920 and a mouse 930.

  In addition, as an embodiment of the present invention, an information providing service for providing information obtained by operating the diagnostic device 400 to an operator and an operation control device equipped with the diagnostic device 400 are also included.

  FIG. 2 is a diagram illustrating the timing of executing the model construction processing mode and the diagnostic mode of the diagnostic apparatus 400. The normal state learning mode (model construction processing mode) is operated every predetermined setting period, and the diagnosis processing mode is operated every sampling period to make a diagnosis.

  FIG. 3 is a diagram illustrating the classification unit 750. In the present embodiment, a case where adaptive resonance theory (ART) is applied as a technique for classifying signals will be described. It should be noted that other clustering methods such as vector quantization and support vector machine can be used.

  As shown in FIG. 3A, the data classification function includes a data preprocessing device 610 and an ART module 620. The data preprocessing device 610 converts the operation data into input data for the ART module 620.

  Hereinafter, those procedures performed by the data preprocessing device 610 and the ART module 620 will be described.

  First, the data preprocessing device 610 normalizes data for each measurement item. Data including the data Nxi (n) obtained by normalizing the measurement signal and the complement CNCN (n) (= 1−Nxi (n)) of the normalized data is defined as input data Ii (n). This input data Ii (n) is input to the ART module 620.

  In the ART module 620, the classification data signal 702, which is input data, or the detection data signal 6 is classified into a plurality of categories.

  The ART module 620 includes an F0 layer 621, an F1 layer 622, an F2 layer 623, a memory 624, and a selection subsystem 625, which are coupled to each other. The F1 layer 622 and the F2 layer 623 are connected via a weighting factor. The weighting coefficient represents a prototype (prototype) of a category into which input data is classified. Here, the prototype represents a representative value of the category.

  Next, the algorithm of the ART module 620 will be described. The outline of the algorithm when input data is input to the ART module 620 is as shown in the following processing 1 to processing 5.

  Process 1: The input vector is normalized by the F0 layer 621, and noise is removed.

  Process 2: A suitable category candidate is selected by comparing the input data input to the F1 layer 622 with the weighting factor.

  Process 3: The validity of the category selected by the selection subsystem 625 is evaluated by the ratio with the parameter ρ. If it is determined to be valid, the input data is classified into the category, and the process proceeds to process 4. On the other hand, if it is not determined to be valid, the category is reset, and an appropriate category candidate is selected from the other categories (repeat process 2). Increasing the value of parameter ρ makes the category classification finer, and decreasing the value of ρ makes the classification coarse. This parameter ρ is referred to as a vigilance parameter.

  Process 4: When all the existing categories are reset in Process 2, it is determined that the input data belongs to the new category, and a new weighting factor representing the prototype of the new category is generated.

Process 5: When input data is classified into category J, weight coefficient WJ (new) corresponding to category J uses past weight coefficient WJ (old) and input data p (or data derived from input data). And updated by the following equation (1).
WJ (new) = Kw.p + (1-Kw) .WJ (old) (1)
Here, Kw is a learning rate parameter (0 <Kw <1), and is a value that determines the degree to which the input vector is reflected in the new weighting factor.

  The arithmetic expressions (1) and expressions (2) to (12) described later are incorporated in the ART module 620.

  The feature of the data classification algorithm of the ART module 620 is the processing 4 described above. That is, in the process 4, when input data different from the learned pattern is input, a new pattern can be recorded without changing the recorded pattern. Therefore, it is possible to record a new pattern while recording a pattern learned in the past.

  As described above, when the operation data given in advance is given as input data, the ART module 620 learns the given pattern. Therefore, when new input data is input to the learned ART module 620, it is possible to determine which pattern in the past is close by the above algorithm. If the pattern has never been experienced before, it is classified into a new category.

FIG. 3B is a block diagram showing the configuration of the F0 layer 621. In F0 layer 621, the input data I again normalized at each time, to create a normalized input vector ^{u 0} is input to the F1 layer 621 and selection subsystem 625.

First, w ⁰ is calculated from the input data I according to the equation (2). Here, a is a constant. In the drawings and formulas, i may be added to these symbols as subscripts.
It is the processing of the second data.

Next, x ⁰ obtained by normalizing w ⁰ is calculated using the equation (3). In yet (3), on both sides of the double vertical lines x ⁰ is a symbol representing the norm.

Then, using the (4) equation, to calculate the ^{v 0} obtained by removing noise from ^{x 0.} However, θ is a constant for removing noise. Since the minute value becomes 0 by the calculation of the equation (4), noise of the input data is removed.

Finally, a normalized input vector u ⁰ is obtained using equation (5). u ⁰ is the input of the F1 layer.

FIG. 3C is a block diagram showing the configuration of the F1 layer 622. As shown in FIG. In F1 layer 622, it holds a ^{u 0} determined by formula (5) as a short-term memory, calculates a pi input to F2 layer 722. Formulas for the F2 layer are collectively shown in formulas (6) to (12). However, a and b are constants, f is a function shown by the equation (4), and Tj is a fitness calculated by the F2 layer 722.

  However,

  FIG. 4 is a diagram for explaining a result of classifying measurement signals or pseudo signals included in the monitoring data information 6 by the classification unit 750.

  FIG. 4A is a diagram for explaining the result of classifying the measurement signal 1 or the pseudo signal acquired from the plant 100 into categories. The horizontal axis represents time, and the vertical axis represents measurement signals and category numbers. According to this result, the items A and B are classified into the category number 1 in the normal period, and then change with time so as to be sequentially classified into the category numbers 2 and 3. Moreover, in the diagnosis period, the state changed so that what was originally classified into category number 2 was classified into category number 4 in a short time.

  FIG. 4B is a diagram illustrating an example of the classification result classified into categories by the classification unit 750. In this example, two items of the measurement signal are displayed and represented by a two-dimensional graph. In addition, the vertical axis and the horizontal axis indicate the measurement signals of the respective items normalized. According to this display result, category number 1 is a group in which item A is large and item B is small, category number 2 is a group in which both item A and item B are small, category number 3 is a group in which item B is large and item A is small, category It can be seen that number 4 is classified into a large group for both item A and item B. Thus, the measurement signal is divided into a plurality of categories 630 (circles shown in FIG. 4B) by the ART module 620 in FIG. One circle corresponds to one category.

  As shown to Fig.4 (a), the data of the normal period before the diagnosis start were classified into the categories 1-3. Data classified into categories 1 to 3 is model data. On the other hand, data in the first half of the diagnosis period after the start of monitoring is classified into category 2, and is data included in the same categories 1 to 3 as the model data. In this case, since the data trends are the same, it is determined that the state has not changed.

  On the other hand, data in the latter half of the diagnosis period after the start of monitoring is classified into category 4, and is classified into a category different from the model data (categories 1 to 3). Since the data trends are different, it is determined that the state of the plant has changed.

  In the present embodiment, an example in which two measurement signals A and B are classified into categories has been described. However, three or more measurement signals can be classified into categories using multidimensional coordinates.

  FIG. 5 is an operation flowchart of the diagnosis mode of the diagnosis apparatus 400.

  In the present invention, when a new category is generated as a result of diagnosis by the state change detection unit 810, it is determined whether or not the driving pattern has changed. When the driving pattern changes, the abnormality detection unit 820 makes a new diagnosis. With this method, it is possible to avoid generation of a false alarm for diagnosing a driving pattern change as abnormal, and at the same time to shorten a period for collecting normal driving data.

  In the operation flowchart of the diagnosis mode of FIG. 5, first, in step 1000, the state change detection unit 810 is operated to make a diagnosis. Next, in step 1010, the diagnosis result in step 1000 is evaluated. If no new category has occurred, the process proceeds to step 1100 (normal determination), and if a new category has occurred, the process proceeds to step 1020.

  In step 1020, the operation pattern evaluation unit 710 is operated to check whether or not the operation pattern has changed. If there is no change, the process proceeds to step 1110 (abnormal determination), and if there is a change, the process proceeds to 1030.

In step 1030, it searches whether a period of the same driving pattern with respect to the measurement signal database 510 to illustrate the current time (time of detecting a state change) in FIGS. 6 (a) is present, the period of the same operation pattern If is not present, the process proceeds to step 1040, if the period of the same operation pattern that exists, the process proceeds to step 1060. In step 1030 in FIG. 5, whether or not there is a period of the same operation pattern as the current time is described as the presence or absence of past data (in the period of the same operation pattern as the current time). For example, in the measurement signal database 510 illustrated in FIG. 6A, when the time 10:01 is the current time and the same operation pattern is searched and this is the time 10:00, the item at this time Data on A, B, and C is extracted as past data.

  In step 1040, whether or not a pseudo signal is to be created is determined based on conditions set in advance by the operator. If a pseudo signal is to be created, the process proceeds to step 1050. If not, the process proceeds to step 1120 (prediction determination). Here, the sign determination is a state where the operation state of the plant is changed but an abnormality does not always occur, and the final determination of the occurrence of the abnormality is performed by the operator.

  In step 1050, first, the similar plant selection unit 720 is operated to extract a plant similar to the diagnosis target plant 100 from the plant database 520. The similar plant selection unit 720 compares the design information of the plant to be diagnosed with the design information of the plant stored in the plant database 520, and extracts a plant that minimizes the evaluation function F of the following equation (13). . In the equation (13), Ci is expressed by the equation (14). Here, as already described, P is an input of the F2 layer 722, and w is a value obtained by equation (2).

F = Σwi × Ci (13)
Ci = (P1i / P2i) ^ 2 (14)
Next, the pseudo signal generation unit 730 creates a pseudo signal, and the process proceeds to step 1060. Details of the pseudo signal generator 730 will be described later with reference to FIG.

  In step 1060, a diagnostic model is constructed using the measurement signal of the same operation pattern extracted from the measurement signal database 510 or the pseudo signal generated in step 1050, and the abnormality detection unit 820 diagnoses using this diagnosis model. . In step 1070, the presence / absence of occurrence of a new category in step 1060 is evaluated.

  After executing steps 1100, 1110, 1120, 1130, and 1140, end determination is performed in step 1080. If there is a request from the plant operator to stop the diagnostic system, the system is terminated. Otherwise, the process returns to step 1000.

  In the diagnostic apparatus of the present invention showing a series of processing flows in FIG. 5, the diagnosis is executed using the diagnostic model using the measurement signal of the same operation pattern in Step 1060, which is compared with the case where the diagnosis is performed in Step 1010. Thus, it is possible to suppress the occurrence of false alarms accompanying the change of driving pattern and improve the diagnostic accuracy. Further, if there is no identical operation data of the diagnosis target plant, a diagnosis signal can be constructed by generating a pseudo signal in step 1050. Therefore, since the operation data of the diagnosis target plant may be a short period, the effect of shortening the time for introducing the diagnosis apparatus can be obtained.

  FIG. 6 is a diagram for explaining an aspect of data stored in the measurement signal database 510, the plant database 520, and the diagnostic model database 530. Here, a state in which these data are described on the monitor screen is shown. Therefore, it is possible to refer to data beyond the display range by appropriately scrolling left and right and up and down on the monitor screen. It is also possible to select a plurality of monitoring groups by tabs.

  FIG. 6A shows the mode of data stored in the measurement signal database 510. Here, the value of the measurement signal 1 (data items A, B, and C are described in the figure), which is operation data measured in the plant 100, is stored for each sampling period (the time on the vertical axis). Further, by using scroll boxes 512 and 513 that can be moved vertically and horizontally on the display screen 511, a wide range of data can be scroll-displayed.

  FIG. 6B shows the mode of data stored in the plant database 520. Here, design values of various plants such as a plant name, output, steam pressure, steam temperature, and the like are stored. In the plant database 520, past operation data of various plants is stored in the format shown in FIG. Furthermore, the plant database 520 stores sensor installation information such as sensor placement positions and the number of measurement points, and environmental information such as atmospheric temperature and humidity.

  FIG. 6C shows a mode of data stored in the diagnostic model database 530. Here, the relationship between the category number and the weighting factor is stored for each diagnostic model (selected by selecting a tab of the monitoring group). Here, the weight coefficient is the center coordinate of the category.

  FIG. 7A is a block diagram illustrating the pseudo signal generation unit 730. The pseudo signal generation unit 730 includes a simulator 732 and a parameter estimation unit 734 as an arithmetic device, and includes parameter data 735 as a memory. The memory may be in the form of a database.

  The simulator 732 uses the information stored in the measurement signal database 510 and the plant database 520 to calculate pseudo data of the measurement signal in the normal state using the following model equation (15).

dx / dt = f (x, α, β) (15)
In the above formula, x is a state quantity of the plant 100, α is an adjustment parameter adjusted by measurement data of a similar plant, and β is a fixed parameter determined by a plant design value stored in the plant database 520.

  The adjustment parameter α is estimated by the parameter estimation unit 734. The present invention is not limited to the parameter estimation method, and various estimation algorithms may be used.

  There is an error between the model formula (15) and the plant characteristic, which is called a modeling error. In the present invention, the modeling error is simulated by assuming that the adjustment parameter α varies within a certain range. In the present embodiment, a Bayesian learning method is used as a method for estimating the representative value and variation range (standard deviation) of the adjustment parameter α.

  Since the Bayesian learning method is described in Non-Patent Document 1, the details of the algorithm are omitted below, and the operation results of the parameter estimation unit 734 and the simulator 732 will be described.

  FIG. 7B shows an example in which an example of parameter data generated in the parameter data 735 is displayed on a display screen. In the attribute column, the adjustment parameter α adjusted with the measurement data of the similar plant and the fixed parameter β determined by the design value of the plant stored in the plant database 520 are classified and expressed, and together with the names of these parameters, The representative value and standard deviation are displayed (stored).

  FIG. 8 is a diagram for explaining the operation results of the parameter estimation unit 734 and the simulator 732. Among these, FIG. 8A is a diagram for explaining parameter sample values obtained as a result of operating the parameter estimation unit 734. In the Markov chain Monte Carlo method of Bayesian estimation, estimated sample values of parameters are generated sequentially. The parameter sample value changes greatly when the number of samplings is small, but decreases as the number of samplings increases. An interval where the number of samplings is a certain value or more is set as an evaluation interval, and an average value and a standard deviation are calculated from sample values of parameters in the evaluation interval.

  FIG. 8B is an example of pseudo data created by the simulator 732. A value when the simulator is operated with the adjustment parameter set to an average value of the sample values is indicated by a solid line, and a value range when the simulator is operated by changing the standard deviation is indicated by a dotted line. In this example, item A is a pseudo data indicating a tendency to remain substantially constant within the range of the upper and lower limit values that are substantially constant, and item B is pseudo data that indicates a tendency to cyclically vary similarly within the range of the upper and lower limit values that periodically vary. I can understand.

  FIGS. 8C and 8D show two items of pseudo data, which are represented by a two-dimensional graph. FIG. 8C is a graph when pseudo data is created using an average value when parameters are changed.

  By changing the parameters, it is possible to generate a pseudo signal that covers the range of values of the measurement signal in consideration of the model error. By constructing a diagnostic model using this pseudo signal, it is possible to suppress the occurrence of false alarms due to model errors. In addition, diagnosis is performed using two diagnostic models: a diagnostic model constructed by classifying pseudo data created with average values, and a diagnostic model constructed by classifying pseudo data when parameters are changed. The result may be output.

  FIG. 9 is a diagram illustrating an example of a screen displayed on the image display device 940. Among these, FIG. 9A is a screen that displays the determination result in the flowchart of FIG. 5, and displays the diagnosis result (either normal determination, abnormal determination, or predictive determination) for each monitoring group.

  FIG. 9B displays the measurement signal and pseudo signal of the similar plant, the result of classifying the pseudo signal into the category, and the design information of the diagnosis target plant and the similar plant. The operation results of the similar plant selection unit 720 and the pseudo signal generation unit 730 can be confirmed on this screen. According to this display screen, when the output of the diagnosis target plant AAAA, the steam pressure, and the steam temperature are 100, 25, and 500, respectively, it is displayed that it is 90, 25, and 520 of the similar plant BBBB.

  FIG. 9C shows a screen used to change the vigilance parameter ρ that determines the classification performance of the category. The effect of changing ρ in FIG. 9C can be evaluated in FIG. 9B. Based on the evaluation result, ρ can be set to an appropriate value.

  FIG. 10 is a diagram showing an embodiment when the plant 100 is a gas turbine plant. FIG. 10A is a diagram illustrating a configuration when the diagnosis target plant 100 is a gas turbine plant. The plant 100 includes a gas turbine generator 110, a control device 120, and a data transmission device 130. The gas turbine generator 110 includes a generator 111, a compressor 112, a combustor 113, and a turbine 114.

  At the time of power generation, the air sucked by the compressor 112 is compressed into compressed air, and this compressed air is sent to the combustor 113 and mixed with fuel and burned. The turbine 114 is rotated using the high-pressure gas generated by the combustion, and the generator 111 generates power.

  In the control device 120, the output of the gas turbine generator 110 is controlled according to the power demand. The control device 120 uses the operation data 102 measured by a sensor (not shown) installed in the gas turbine generator 110 as input data. The operation data 102 is state quantities such as intake air temperature, fuel input amount, turbine exhaust gas temperature, turbine rotation speed, generator power generation amount, turbine shaft vibration, and the like, and is measured at each sampling period. It also measures weather information such as atmospheric temperature.

  In the control device 120, the control signal 101 for controlling the gas turbine generator 110 is calculated using these operation data 102. In addition, the control device 120 performs processing for generating an alarm when the value of the operation data 102 deviates from a preset range. The alarm signal is processed as a digital signal of 1 when the operation data 102 deviates from a preset range, and 0 when within the range. When the alarm signal is 1, the operator is notified of the content of the alarm by sound or screen display.

  The signal data transmission device 130 transmits the operation data 102 measured by the control device 120, the control signal 101 calculated by the control device 120, and the measurement signal 1 including the alarm signal to the plant state monitoring device 200.

  As shown in FIG. 10B, there are a startup mode A and a control mode B as variations of the operation pattern at the start-up in this plant. The activation mode A is an activation method according to the heat insulation state of the device, and includes modes such as hot start, warm start, and cold start. The control mode B is a control method that is determined according to the control purpose, and includes modes such as the shortest start-up time control, the minimum life consumption control, and the maximum power generation amount control during start-up.

  When the present invention is realized, the operation pattern evaluation unit 710 in FIG. 1 may distinguish the operation pattern according to the mode to which the activation mode A and the control mode B belong. Needless to say, there are no restrictions on the type of mode and the contents of each mode in carrying out the present invention.

  FIG. 11 is a diagram showing an embodiment when the plant 100 is a boiler plant. Fig.11 (a) is a figure which shows a structure in case the plant 100 of a diagnostic target is a boiler plant. This plant is a thermal power plant having a boiler 200 and a steam turbine 300 driven by steam generated by the boiler 200 as main components (a generator is not shown). The boiler plant 100 controls the specified load (power generation output) based on the load request command. By adjusting the valve opening degree of the steam control valve 290, the steam flow rate 261 guided to the turbine 300 changes and the power generation output changes.

  In addition, the water / steam system includes a condenser 310 that cools the steam emitted from the steam turbine 300 into a liquid, and a water supply pump 320 that feeds water cooled by the condenser 310 back to the boiler 200 as boiler feed water. There is. Although not shown in the drawings, an actual plant also includes a feed water heater that preheats boiler feed water using a part of steam extracted from the middle stage of the steam turbine 300 as a heating source.

  On the other hand, in the system of the combustion gas 201 discharged from the boiler, there are an exhaust gas treatment device 330 for purifying exhaust gas and a chimney 340 for discharging the purified gas 331.

  Coal 381 as fuel is sent to a coal pulverizer (mill) 350 via a fuel supply amount adjustment valve 380. Air 382 used for coal conveyance and combustion adjustment is supplied to the coal pulverizer 350 and the burner 310 via the air amount adjustment valve 370. Coal that has been powdered (pulverized coal) by the coal pulverizer 350 is conveyed by air and supplied to the burner 210. An after air port 220 is disposed on the upper portion of the burner 210, and air 383 is supplied to the after air port 220 via an air amount adjustment valve 360.

  Next, the configuration of the boiler 200 will be described. A furnace having a burner 210 for burning fuel has a cooling wall called a water wall 230 that cools the entire wall surface and collects the heat of the combustion gas because the inside of the furnace becomes hot. In the boiler 200, there are other heat exchangers including a economizer 280, a primary superheater 270, a secondary superheater 240, a tertiary superheater 250, and a fourth superheater 260. Is recovered to produce high-temperature steam.

  Although not shown in the figure, the plant has many sensors for measuring gas composition, temperature, pressure, steam temperature, pressure, heat exchanger metal temperature, etc. The measurement result is transmitted as measurement information 1 from the data transmission device 390 to the boiler tube leak detection device 400. Although not shown, it is common to install a plurality of boiler burners 210 in the horizontal direction before and after the furnace and a plurality of stages in the height direction, and a plurality of after airports in the horizontal direction before and after the furnace. is there.

  As shown in FIG. 11B, there are a burner pattern, a fuel type mode, and a control mode as variations of the operation pattern in this plant.

  A plurality of boiler burners 210 are installed in the horizontal direction and in a plurality of stages in the height direction before and after the furnace, and the number of burners to be ignited is determined according to the output of the plant. Since the position of the burner to be ignited can be set arbitrarily, there are a plurality of combinations of ignition burners. Moreover, the component of coal changes with brands of coal used with a boiler. If the burner pattern and fuel type are different, the heat absorption characteristics of the furnace will change, which affects the plant characteristics such as steam temperature. The control mode is a control method that is determined according to the control purpose, and includes modes such as minimum fuel consumption control and minimum NOx control.

  When realizing the present invention, the operation pattern evaluation unit 710 in FIG. 1 distinguishes operation patterns according to the burner pattern mode A, the fuel type mode B, and the control mode C. Needless to say, there are no restrictions on the type of mode and the contents of each mode in carrying out the present invention.

  FIG. 12 is a block diagram for explaining a fourth embodiment of the present invention.

  In this embodiment, a data monitoring center 950 and a site 970 are connected to each other via an information communication network 960. Although there are a plurality of power plants 970 and three sites are described in the present embodiment, any number of sites may be connected. The communication information 20 communicated between the sites includes information on the model formula and the adjustment parameter α. The data monitoring center 950 confirms that the communication information 20 does not include plant design information and measurement signals.

  In the plant database 520 provided in the diagnosis apparatus 400 described in the first embodiment, design information and measurement signals of other plants are stored in addition to the plant design information to be diagnosed. However, it is generally difficult to obtain design information and measurement signals of other plants. In addition, management of such information requires a large amount of cost for preventing information security accidents.

  In the present embodiment, although not shown, plant design information to be diagnosed is stored in the plant database 520 provided in the diagnostic apparatuses 400a to 400c, and design information and measurement signals of other plants are not stored. The pseudo signal is calculated by the equation (4) using the model equation and the adjustment parameter value included in the communication information 12 communicated between the sites.

  In this embodiment, it is possible to operate the diagnostic apparatus at each site without providing plant design information and measurement signal information to other sites, and the effect of enhancing information security can be obtained.

  The present invention is applicable to various plants such as a thermal power plant and a nuclear power plant as a diagnostic device.

1: Measurement signal 2: External input signal 3: Measurement signal 4: Measurement signal 5: Plant database information 6: Monitoring data information 7: Classification result information 8: Diagnostic model database information 9: Diagnostic model database information 10: Detection result information 11 : Image display information 100: Plant 400: Diagnosis device 410: External input interface 420: External output interface 510: Measurement signal database 520: Plant database 530: Diagnosis model database 700: Monitoring data generation unit 710: Operation pattern evaluation unit 720: Similar Plant selection unit 730: pseudo signal generation unit 800: detection unit 810: state change detection unit 820: abnormality detection unit 900: operation management room 910: external input device 920: keyboard 930: mouse 940: image display device

Claims (11)

A diagnostic device for a plant that includes a detection unit that diagnoses the operating state of the plant using a measurement signal obtained by measuring a state quantity of the plant, and displays a diagnostic result on an image display device,
A measurement signal database for storing measurement signals obtained by measuring state quantities of the plant to be diagnosed;
It has a plant database that stores past operation data of various plants,
The detector is
A state change detection unit that detects a state change using a diagnosis model constructed with data of a predetermined period;
About the operation pattern that is an external environmental factor that is not directly related to the device characteristics, an abnormality detection unit that detects presence or absence of abnormality using a diagnosis model constructed with data of the same operation pattern as when the state change was detected,
The data of the same operation pattern uses the measurement signal of the diagnosis target plant stored in the measurement signal database , or when the measurement signal of the diagnosis target plant stored in the measurement signal database cannot be used, A plant diagnosis apparatus comprising a pseudo signal generation unit that calculates from past operation data of a similar plant selected from plants other than the diagnosis target plant . The plant diagnostic apparatus according to claim 1 ,
The pseudo signal generation unit is a simulator that is a model that generates a pseudo signal that simulates the characteristics of a diagnosis target plant using adjustment parameters and fixed parameters, and a function that determines the adjustment parameters using operation data of similar plants. And a parameter estimation unit that determines the fixed parameter based on design information of the plant to be diagnosed. The plant diagnostic apparatus according to claim 2 ,
The parameter estimation unit estimates a representative value and a standard deviation of parameter values for the adjustment parameter, and the simulator generates a pseudo signal by changing the adjustment parameter within a standard deviation range. Plant diagnostic equipment. A plant diagnostic apparatus according to any one of claims 1 to 3,
In the state change detection unit, when using a diagnostic model constructed with data of the predetermined period, the predetermined period is determined according to an operation cycle of the plant. The plant diagnostic apparatus according to claim 2,
A plant diagnosis apparatus, comprising: an image display device that displays measurement signals and pseudo signals of the similar plant, results of classifying the pseudo signals into categories, and design information of the diagnosis target plant and the similar plant. A plant diagnostic apparatus according to any one of claims 1 to 5 , comprising:
The plant diagnosis apparatus according to claim 1, wherein the plant is a gas turbine plant, and the operation pattern is determined by a start mode corresponding to a heat insulation state of the gas turbine or a control mode determined according to a control purpose. A plant diagnostic apparatus according to any one of claims 1 to 5 , comprising:
The plant is a thermal power plant including a boiler, and the operation pattern is determined by a boiler burner pattern, a fuel type mode, or a control mode. A plant diagnostic apparatus according to claim 2 or claim 3 , wherein
The plant diagnosis apparatus characterized in that the value of the adjustment parameter is transmitted and received via a network. A plant diagnosis method for diagnosing the operation state of a plant using a measurement signal obtained by measuring a state quantity of the plant and displaying a diagnosis result,
Memorize the measurement signal that measured the state quantity of the plant to be diagnosed,
Stores past operation data of various plants,
Detects state changes using a diagnostic model built with data for a given period,
For driving patterns that are external environmental factors that are not directly related to device characteristics, the presence or absence of abnormalities is detected using a diagnostic model built with the same driving pattern data as when the state change was detected.
The data of the same operation pattern uses the stored measurement signal of the diagnosis target plant , or if the stored measurement signal of the diagnosis target plant cannot be used, select from the plants other than the diagnosis target plant A method for diagnosing a plant characterized in that it is obtained from past operation data of a similar plant. A plant diagnosis method for diagnosing the operation state of a plant using a measurement signal obtained by measuring a state quantity of the plant and displaying a diagnosis result,
Obtain a category of the measurement signal using a diagnostic model constructed with data for a predetermined period, and if a new category occurs, check whether the operation pattern of the plant is an external environmental factor that is not directly related to the equipment characteristics. Check whether there is a period of the same operation pattern as the time when the new category is detected, and if there is a period of the same operation pattern , build a diagnostic model using the measurement signal in the operation pattern of the period , While evaluating the occurrence of a new category in the diagnostic model to determine the abnormality ,
When a period of the same operation pattern does not exist, a diagnostic model is constructed using a pseudo signal created using design information of a plant similar to the plant to be diagnosed, and the occurrence of a new category in the diagnostic model is evaluated. A method for diagnosing a plant, wherein abnormality determination is performed. The plant diagnosis method according to claim 10 , comprising:
The pseudo signal is calculated using a simulator that is a model that generates a pseudo signal that simulates the characteristics of the plant to be diagnosed using the adjustment parameter and the fixed parameter, and the adjustment parameter is calculated using operation data of a similar plant. A plant diagnosis method characterized by determining and determining the fixed parameter based on design information of a plant to be diagnosed.

Images