CN102999038B - Power generation facility diagnostic device and power generation facility diagnostic method - Google Patents

Abstract

A diagnostic device reduces the false negative that an abnormality cannot be detected when the abnormality occurs, and has high diagnostic accuracy. A diagnostic device for a power generating plant, which diagnoses the operating state of the plant based on a measurement signal obtained by measuring a state quantity from the power generating plant and displays the diagnosis result on an image display device, is provided with: a model building unit for building a model used for diagnosis by using a measurement signal obtained by measuring a state quantity of the power generation equipment in a diagnosis device of the power generation equipment; a model definition unit that defines an operating condition for diagnosis using the model and a normalization method of a measurement signal; a diagnosing unit for diagnosing an operation state of the power generating equipment using the model constructed by the model constructing unit; the model definition means includes an operating condition determination unit that determines an operating condition of the power generation plant and a normalization condition determination unit that determines a normalization condition of the data for each operating condition determined by the operating condition determination unit, and the diagnosis means switches the diagnostic model to perform diagnosis in accordance with the operating condition.

Description

Diagnosis device for power generation equipment, and diagnosis method for power generation equipment

technical field

The present invention relates to a diagnosis device for a power generation plant and a diagnosis method for a power generation plant.

Background technique

The diagnosis device of the power generation equipment detects the occurrence of the abnormality or the accident based on the measurement signal from the equipment when an abnormal transient phenomenon or an accident occurs in the equipment.

As a known example of a device diagnostic device, JP 2005-165375 A discloses a device diagnostic device using Adaptive Resonance Theory (ART). Here, ART is a technique for classifying multidimensional data into categories according to their similarity.

In the technology of the device diagnosis device described in Japanese Unexamined Patent Application Publication No. 2005-165375, first, the measurement signal during the period when the state of the device is considered to be normal is extracted from the past measurement signal in which the operation data of the device is recorded, and the as data for model building. Then, using ART, the data for model construction is classified into a plurality of categories (normal categories), and a normal model used for diagnosis is created. Next, ART is used to classify the current measurement signals of the device into categories. When the current measurement signal does not match the normal model, that is, when it cannot be classified into the normal class, a new class (new class) is generated. That is, the creation of a new category means that the tendency of the measurement signal has changed, and the state of the device has changed. Therefore, this is a technique for judging the occurrence of an abnormality by the occurrence of a new category, and diagnosing it as an abnormality when the rate of occurrence of a new category exceeds a threshold value.

patent documents

Patent Document 1: JP-A-2005-165375

Power generation equipment starts, stops, constant load, load changes, etc. are operated under various conditions. If the operating conditions are different, the change range of the measurement signal is also different.

In addition, normalization processing is used as preprocessing of measurement signals used for diagnosis. In normalization processing, the lower limit value of normalization is set to 0, and the upper limit value of normalization is set to 1, so as to process the measurement signal. The normalized lower limit and upper limit need to be set in advance. In prior art diagnostic devices, the measurement signals are processed under the same normalization conditions regardless of the operating conditions. Therefore, it is necessary to set the normalization range to a wide range so as to cover the variation range of the measurement signal under all operating conditions.

If the normalization range is larger than the variation range of the measurement signal, the variation in the value of the measurement signal after normalization becomes smaller. Therefore, it is impossible to capture the tendency change of the measurement signal when an abnormality occurs, and there is a possibility that a new category cannot be generated even when an abnormality occurs. This became the cause of underreporting.

Contents of the invention

An object of the present invention is to provide a diagnostic device that appropriately determines a normalization range in accordance with operating conditions, thereby suppressing false positives and improving diagnostic accuracy.

A diagnosis device for power generation equipment, which diagnoses the operating state of the equipment based on a measurement signal obtained by measuring a state quantity from the power generation equipment, and displays the diagnosis result on an image display device, is characterized in that it includes: a model construction unit, which is used in power generation equipment The measurement signal obtained by measuring the state quantity of the power generation equipment in the diagnosis device of the equipment is used to construct the model used in the diagnosis; the model definition unit defines the operating conditions and the normalization method of the measurement signal for diagnosis by the model; and a diagnosis unit that diagnoses the operating state of the power generation facility using the model constructed by the model construction unit; the model definition unit includes: an operation condition judging unit that judges the operation condition of the power generation facility; and a normalization condition The determining unit determines a normalization condition of the data for each operating condition determined by the operating condition determining unit, and the diagnosis unit performs diagnosis by switching a diagnosis model according to the operating condition.

By using the diagnostic device for power generation equipment of the present invention, it is possible to reduce the false alarm that the abnormality cannot be detected when the abnormality occurs, and improve the diagnostic accuracy. In addition, the normalization range can be automatically determined, and the adjustment period of the diagnostic device can be shortened.

Description of drawings

FIG. 1 is a control block diagram showing the configuration of a diagnostic device for power generation equipment according to an embodiment of the present invention.

FIG. 2 is a flowchart showing basic operations of the diagnostic device for power generation equipment shown in FIG. 1 and an explanatory diagram showing timings of operations.

3 is an explanatory diagram showing an implementation example of a function of classifying data in a model construction unit and a diagnosis unit in the diagnostic device for equipment shown in FIG. 1 .

FIG. 4 is an explanatory diagram showing an example of classifying measurement signals by a model construction unit in the diagnostic device for power generation equipment shown in FIG. 1 .

Fig. 5 is an explanatory diagram showing a relationship between a start pattern and a process value in the power generation facility shown in Fig. 4(a).

FIG. 6 is an explanatory diagram of the operation of the operating condition determination unit 500 in the diagnostic device for power generation equipment shown in FIG. 1 .

FIG. 7 is an explanatory diagram of a first embodiment of the normalization condition determination unit 600 in the diagnostic device for power generation equipment shown in FIG. 1 .

FIG. 8 is an explanatory diagram of a second embodiment of the normalization condition determination unit 600 in the diagnostic device for power generation equipment shown in FIG. 1 .

FIG. 9 is an explanatory diagram of a third embodiment of the normalization condition determination unit 600 in the diagnostic device for power generation equipment shown in FIG. 1 .

FIG. 10 is an explanatory diagram of an operation flow in a model construction mode and a diagnosis mode in the diagnostic device for power generation equipment shown in FIG. 1 .

FIG. 11 is an explanatory diagram of a data format stored in a database in the diagnostic device for power generation equipment shown in FIG. 1 .

Fig. 12 is an explanatory diagram of an application effect of the diagnostic device for power generation equipment shown in Fig. 1 .

FIG. 13 is an explanatory diagram of a screen displayed on an image display device in the diagnostic device for power generation equipment shown in FIG. 1 .

Symbol Description:

1, 3, 4 measurement signal

2 external input signal

5, 6, 7 Model Definition Information

8, 9 Model Information

10Diagnostic results

11 screen display information

50 diagnostic device information

100 devices

200 diagnostic devices

210 external input interface

220 external output interface

310 measurement signal database

320 model definition database

330 Diagnostic Model Database

400 model definition units

500 operating condition judgment department

600 Normalized condition decision department

700 model building units

800 diagnostic unit

900 Operation Management Room

910 external input device

920 keyboard

930 mouse

940 image display device

Detailed ways

Next, a diagnostic device for a power generation facility as an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram illustrating a diagnosis device of a power generation facility as an embodiment of the present invention. In the diagnostic device for power generation equipment shown in FIG. 1 , the state of the equipment 100 is diagnosed by the diagnostic device 200 .

The diagnostic device 200 includes a model definition unit 400 , a model construction unit 700 , and a diagnostic unit 800 as computing devices constituting the diagnostic device 200 . The diagnosis device 200 includes a measurement signal database 310 , a model definition database 320 , and a diagnosis model database 330 as databases. In addition, in FIG. 1, the database is abbreviated as DB.

Electronic information is recorded in the databases of the measurement signal database 310, the model definition database 320, and the diagnosis model database 330, and these are generally called electronic files (electronic data).

The model constructing unit 700 creates a diagnostic model that learns the normal state of the device based on the accumulated data of the measurement signal of the past operation state of the device 100 based on the measurement signal of the operating state of the measuring device 100 .

The model definition unit 400 defines the operating conditions for diagnosis by the diagnosis model and the normalization method of the measurement signal when the model is constructed.

The diagnostic unit 800 compares the value data of the diagnostic model created by the model construction unit 700 with the measured data of the measurement signal of the device 100 . If the measurement signal matches the value of the diagnostic model that has learned the normal state, the state of the device is determined to be normal, and if it does not match, it is determined to be abnormal.

In addition, the diagnostic device 200 includes an external input interface 210 and an external output interface 220 as interfaces with the outside.

Then, through the external input interface 210, the operating state of the equipment 100, that is, the measurement signal 1 obtained by measuring various state quantities, and the operation of the external input device 910 composed of a keyboard 920 and a mouse 930 provided in the operation management room 900 The generated external input signal 2 is taken into the diagnostic device 200 . 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 220 .

In addition, in the device diagnosis device of the present embodiment, the diagnosis device 200 includes the model definition unit 400, the model construction unit 700, the diagnosis unit 800, the measurement signal database 310, the model definition database 320, and the diagnosis model database 330. However, Some of them may be arranged outside the diagnostic device 200, and only data may be communicated between these devices.

In addition, a case is shown in which one device 100 to be diagnosed is one in the device diagnostic device of this embodiment, but a single diagnostic device 200 may be used to diagnose a plurality of devices 100 .

Next, the operation of the diagnostic device 200 included in the diagnostic device of the device of this embodiment will be described.

In the device diagnosis device of this embodiment shown in FIG. 1 , measurement signals 1 obtained by measuring various state quantities of the device 100 are taken in via the external input interface 210 . The measurement signals 3 are stored in a measurement signal database 310 provided in the diagnostic device 200 .

The model definition unit 400 includes an operating condition determination unit 500 and a normalization condition determination unit 600 , respectively. The model definition unit 400 outputs the model definition information 5 to the model definition database 320 in response to the input of the measurement signal 4 .

The power generation facility has a constant load operation in which the output is kept constant, a startup operation in which the facility is started, a shutdown operation in which the facility is stopped, and a load variation operation in which the output is varied. The operation condition determination unit 500 uses the accumulated data of the equipment 100 accumulated in the measurement signal database 310 and divides these data into operation patterns of constant load operation, startup operation, shutdown operation, and load variation operation. In addition, the feature quantities of the respective operation modes are extracted. The details of this function will be described using FIG. 6 .

In addition, in the model constructing unit 700, the measurement signal is subjected to normalization processing as pre-processing for constructing a model. In the normalization condition determination unit 600, an appropriate normalization condition is determined for each operation mode. The details of this function will be described later using FIGS. 7 to 9 . The aforementioned model definition information 5 is composed of operating condition data and normalization condition data.

In model constructing unit 700 , a model used for diagnosis is constructed using measurement signal 4 of device 100 stored in measurement signal database 310 and model definition information 6 stored in model definition database 320 . The model information 8 created by the model construction unit 700 is stored in the diagnosis model database 330 .

As techniques for implementing the model construction unit 700, there are clustering techniques such as adaptive resonance theory and vector quantization. In addition, the model used for the diagnosis is not limited to the above-mentioned clustering method, and a statistical model using a physical model, a neural network, or the like can also be used.

In the diagnosis unit 800 provided in the diagnosis apparatus 200, with respect to the input of the measurement signal 4, by referring to the model definition information 7 of the model definition database 320 and the model information 9 of the diagnosis model database 330, the device 100 is evaluated. Diagnose the running state and output the diagnosis result 10.

The diagnostic result 10 diagnosed by the diagnostic unit 800 with respect to the current operating state of the equipment 100 is sent to the image display device 940 provided in the operation management room 900 via the external output interface 220 as image display information 11 and displayed. show. As a result, the operator located in the operation management room 900 is notified of the diagnosis result for the operating state of the equipment 100 .

In this manner, in the device diagnosis device 200 of this embodiment, the operator is notified of the change in the state of the device.

In addition, the diagnostic device information 50 stored in the measurement signal database 310 , the model definition database 320 , and the diagnostic model database 330 installed in the diagnostic device 200 can be arbitrarily displayed on the image display device 940 of the operation management room 900 . In addition, these pieces of information can be corrected by operating the external input signal 2 generated by operating the external input device 910 including the keyboard 920 and the mouse 930 .

Next, the operation of the diagnosis device of the equipment of this embodiment will be described. Next, an operation flowchart of the diagnosis device 200 will be described using FIG.

As shown in the flowchart of FIG. 2( a ), the basic operation of the diagnostic device 200 is executed by combining steps 201 , 202 , and 203 .

First, in step 201, it is determined whether the operation mode of the diagnostic device 200 is the model building mode or the diagnosis mode. Then, the process proceeds to step 202 in the case of the model construction mode, and proceeds to step 203 in the case of the diagnosis mode.

When step 202 is operated, the model definition unit 400 and the model construction unit 700 operate. As a result, model definition information 5 and model information 8 are generated, and the generated information is stored in model definition database 320 and diagnostic model database 330 , respectively. The details of the operation in the model building mode will be described later using FIG. 10( a ).

In addition, if step 203 is activated, the diagnosis unit 800 diagnoses the operating state of the equipment 100, and by sending the image display information 11 including the diagnosis result to the image display device 940, the image display device 940 displays the status of the equipment 100. Operating status. The details of the operation in the diagnosis mode will be described later using FIG. 10( b ).

Timings for operating the model building mode and the diagnosis mode of the diagnostic device 200 can be arbitrarily designated by the operator. Next, various examples of timings for operating the model construction mode and the diagnosis mode will be described with reference to FIGS. 2( b ) to ( d ).

In the embodiment shown in FIG. 2( b ), diagnosis is performed by operating both the model building mode and the diagnosis mode for every sampling period of the measurement signal.

By updating the diagnosis model every time a measurement signal is acquired, it is possible to perform diagnosis using the latest model at all times.

However, when the amount of data used for model construction is large, the calculation may not be completed within the sampling period because the model construction takes time.

In such a case, as in the embodiment shown in FIG. 2(c), it is also possible to activate the normal state model construction mode for each predetermined setting period, and only activate the diagnosis mode for each sampling period, thereby achieving Make a diagnosis. In the method of the embodiment shown in FIG. 2( b ) and FIG. 2( c ), the diagnosis mode is executed every sampling period, and the state of the device can be diagnosed online.

In addition, as in the embodiment shown in FIG. 2( d ), by inputting an external input signal 2 for performing model construction and diagnosis to the diagnosis device 200 by the operator, the model construction mode and the diagnosis mode can be activated at any timing. . That is, various conditions can be changed to diagnose the operating state of the device 100 .

Next, the function of classifying the measurement signal 4 of the equipment 100 included in the model constructing unit 700 and the diagnosing unit 800 of the diagnostic device 200 constituting the diagnostic device for equipment of this embodiment will be described with reference to FIGS. 3 and 4 .

In the diagnosis device of the device of the present embodiment, a case where Adaptive Resonance Theory (Adaptive Resonance Theoty: ART) is used for the data classification function will be described. In addition, as a data classification function, other clustering methods such as vector quantization can be used.

As shown in FIG. 3( a ), the data classification function is composed of a data preprocessing device 710 and an ART module 720 . The data preprocessing device 710 converts the operating data into input data of the ART module 720 .

Next, the procedures (processes) performed by the data preprocessing device 710 and the ART module 720 will be described.

First, in the data preprocessing device 710 , the data is normalized for each measurement item using the information of the normalization condition stored in the model definition database 320 . Let the data including the normalized data Nxi(n) of the measurement signal and the complement CNxi(n) (=1−Nxi(n)) of the normalized data be input data Ii(n ). The input data Ii(n) is input into the ART module 720 .

In the ART module 720 , the measurement signals 4 of the device 100 as input data are classified into a plurality of classes.

The ART module 720 includes an F0 layer 721 , an F1 layer 722 , an F2 layer 723 , a memory 724 , and a selection subsystem 725 , which are integrated with each other. The F1 layer 722 and the F2 layer 723 are combined via weight coefficients. The weight coefficients represent prototypes of classes that classify the input data. Here, a prototype is a type representing a representative value of a category.

Next, the algorithm of the ART module 720 will be described.

The outline of the algorithm in the case of inputting the input data to the ART module 720 is as in the following processing 1 to processing 5 .

Process 1: Normalize the input vector through the F0 layer 721 to remove noise.

Process 2: By comparing the input data input to the F1 layer 722 with the weighting coefficients, suitable category candidates are selected.

Process 3: Evaluate the rationality of the category selected by the selection subsystem 725 by the ratio to the parameter ρ. If judged to be reasonable, the input data is classified into the category, and the process proceeds to step 4 . On the other hand, if it is not determined to be reasonable, the category is reset, and a suitable category candidate is selected from other categories (repeat process 2). If the value of the parameter ρ is made larger, the classification of the categories becomes finer, and if the value of the parameter ρ is made smaller, the classification becomes rough. This parameter ρ is called a vigilance parameter.

Process 4: If all known categories are reset in Process 2, it is determined that the input data belongs to a new category, and a new weight coefficient representing a prototype of the new category is generated.

Processing 5: If the input data is classified into category J, the weight coefficient WJ (new: new) corresponding to category J uses the past weight coefficient WJ (old: old) and the input data p (or derived from the input data data) is updated according to the following formula (1).

[number 1]

WJ(new)＝Kw·p+(1-Kw)·WJ(old)...Formula (1)

Here, Kw is a learning rate parameter (0<Kw<1), and is a value that determines the degree to which an input vector is reflected in a new weight coefficient.

In addition, the ART module 720 incorporates Expression (1) and each calculation expression of Expression (2) to Expression (12) described later.

The data classification algorithm of the ART module 720 is characterized by Process 4 described above.

In process 4, when input data different from the pattern (pattern) at the time of learning is input, a new pattern can be recorded without changing the already recorded pattern. Therefore, new patterns can be recorded while recording patterns learned in the past.

In this way, when the operation data given in advance as input data is given, the ART module 720 learns the given pattern. Therefore, when new input data is input to the ART module 720 that has completed learning, it can be determined which pattern the input data is close to in the past by the above-mentioned algorithm. In addition, if a type has not been experienced in the past, it is classified into a new category.

FIG. 3( b ) is a block diagram showing the structure of the F0 layer 721 . In the F0 layer 721, the input data I _i is normalized again at each time point, and the normalized input vector u _i ⁰ input to the F1 layer 721 and the selection subsystem 725 is created.

First, according to the input data I _i , w _i ⁰ is calculated according to formula (2), where a is a constant.

[number 2]

w _i ⁰ ＝I _i +au _i ⁰ ...Formula (2)

Next, formula (3) is used to calculate x _i ⁰ obtained after normalizing W _i ⁰ . Here, ||·|| is a symbol representing a norm (norm).

[number 3]

${x_i}^0=\frac{{x_i}^0}{\left|\right|w^0\left|\right|}$ ...Formula (3)

Then, using Equation (4), calculate V _i ⁰ after removing noise from _xi ⁰ . where θ is a constant used to remove noise. According to the calculation of the formula (4), since the small value becomes 0, the noise of the input data is removed.

[number 4]

${v_i}^0=f\left({x_i}^0\right)=\left\{\begin{array}{cc}{x_i}^0& \mathrm{if}{x_i}^0\&Greater\; Equal;\mathrm{\&theta;}\\ 0& \mathrm{otherwise}\end{array}\right.$ ...Formula (4)

Finally, formula (5) is used to obtain the normalized input vector u _i ⁰ . u _i ⁰ is the input of F1 layer.

[number 5]

${u_i}^0=\frac{{v_i}^0}{\left|\right|v^0\left|\right|}$ ...Formula (5)

FIG. 3( c ) is a block diagram showing the configuration of the F1 layer 722 . In the F1 layer 722, u _i ⁰ obtained by the expression (5) is held as short-term storage, and p _i input in the F2 layer 722 is calculated. The calculation formulas for the F2 layer are collectively shown in formula (6) to formula (12). Here, a and b are constants, f(·) is a function represented by the formula (4), and T _j is the suitability calculated in the F2 layer 722 .

"Number 6]

w _i ＝u _i ⁰ +au _i ...Formula (6)

[number 7]

$x_i=\frac{w_i}{\left|\right|w\left|\right|}$ ...Formula (7)

[number 8]

v _i = f(x _i )+bf(q _i )...Formula (8)

[Number 9]

$u_i=\frac{v_i}{\left|\right|v\left|\right|}$ ...Formula (9)

[number 10]

$q_i=\frac{p_i}{\left|\right|p\left|\right|}$ ...Formula (10)

[number 11]

$p_i=u_i+\underset{i}{\overset{m}{\mathrm{\&Sigma;}}}g\left({\mathrm{the\; y}}_i\right)z_{\mathrm{the\; ji}}$ ...Formula (11)

in,

[number 12]

$g\left({\mathrm{the\; y}}_i\right)=\left\{\begin{array}{cc}d& \mathrm{if}{T}_j=\max \left(T_j\right)\\ 0& \mathrm{otherwise}\end{array}\right.$ ...Formula (12)

Next, the function of constructing a model using the measurement signal 4 of the device 100 included in the model constructing unit 700 of the diagnostic device 200 constituting the diagnostic device of the device according to the present embodiment will be described using FIG. 4 .

First, an embodiment of the device 100 will be described using FIG. 4( a ), and the information contained in the measurement signal 4 will be described. Next, how the measurement signal 4 is classified into categories will be described using FIG. 4( b ) and FIG. 4( c ).

FIG. 4( a ) is a block diagram showing a thermal power generation facility which is an example of the facility 100 .

In FIG. 4( a ), a thermal power generation facility 100 includes a gas turbine generator 110 , a control device 120 , and a data transmission device 130 . Gas turbine generator 110 includes generator 111 , compressor 112 , combustor 113 , and turbine 114 .

When power generation is performed, the air taken in by the compressor 112 is compressed to form compressed air, and this compressed air is sent to the combustor 113, where it is mixed with fuel and burned. The turbine 114 is rotated using high-pressure gas generated by combustion, and the generator 111 generates electricity.

In the control device 120, the output of the gas turbine generator 110 is controlled according to the electric power demand. In addition, the control device 120 uses the operating data 102 measured by a sensor (not shown) provided in the gas turbine generator 110 as input data. The operating data 102 are state quantities such as intake air temperature, fuel input amount, turbine exhaust temperature, turbine rotational speed, generator power generation, turbine shaft vibration, etc., and are measured every sampling cycle. In addition, weather information such as atmospheric temperature is also measured.

In the control device 120 , the control signal 101 for controlling the gas turbine generator 110 is calculated using these operating data 102 . Moreover, in the control apparatus 120, the process which issues an alarm when the value of the operation data 102 deviates from the preset range is also performed. The alarm signal is processed as a digital signal that is "1" when the operating data 102 is out of a preset range and is "0" when it is within the range. When the alarm signal is "1", the content of the alarm is notified to the operator by sound, screen display, or the like.

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

FIG. 4( b ) is a diagram illustrating the result of classifying the measurement signal 1 acquired from the device 100 into categories. The horizontal axis is time, and the vertical axis is measurement signal and category number. FIG. 4( c ) is a diagram showing an example of a classification result of classifying the measurement signal 1 of the device 100 into categories.

FIG. 4( c ) shows two items in the measurement signal as an example, and is marked with a two-dimensional graph. In addition, the vertical axis and the horizontal axis represent the normalized measurement signals of the respective items.

The measurement signal is divided into a plurality of categories 1000 (circles shown in FIG. 4( c )) by the ART module 720 of FIG. 3( a ). A circle corresponds to 1 category.

In this embodiment, measurement signals are classified into 4 categories. Category number 1 indicates a group with a large value for item A and a small value for item B, category number 2 indicates a group with a small value for both item A and item B, and category number 3 indicates a group with a small value for item A and a small value for item B The group with a large value of , and the category number 4 indicates a group with both item A and item B with large values.

As shown in FIG. 4( b ), the data of the normal period before the start of the diagnosis are classified into categories 1-3. The data in the first half after the start of the diagnosis is classified into category 2, which is the same category as the normal period. In this case, since the trend of the data is the same as the normal period, it is diagnosed as normal. On the other hand, the data in the second half after the start of diagnosis is classified into category 4, which is a different category from the normal period. Since the tendency of the data is different, there is a possibility that the state of the equipment changes and an abnormality occurs. In this case, in the diagnostic device 200 of the present invention, the possibility of abnormality is displayed on the image display device 940 to the operator of the facility, and the operator is notified.

In addition, in this embodiment, an example of classifying measurement signals of two items into categories is described, but measurement signals of three or more items may be classified into categories using multidimensional coordinates.

Fig. 5 is a diagram illustrating a relationship between a start pattern and a process value in the power generation plant shown in Fig. 4(a), showing temporal changes in an output command value and a turbomachinery temperature.

Typical start modes include hot start and cold start. Restarting while the turbine and compressor are hot is called hot start. In addition, restarting in a state in which the turbine and the compressor have cooled down after stopping for a relatively long time is called a cold start. As shown in FIG. 5( b ), if the starting mode is different, the temperature range of the turbo equipment at the start of starting and the temperature during the load change are also different.

In a conventional diagnostic device, one diagnostic model is constructed regardless of operating conditions, so it is necessary to determine a normalization range to include data. That is, for example, the upper limit value of the normalization range is set to 1010, and the lower limit value is set to 1011.

For example, the normalized range 1022 is widened compared to the range 1020 over which the process value varies during a warm start at low load.

If the normalization range is wider than the variation range of the data, the variation of the normalized value becomes smaller. Therefore, changes in data at the time of an abnormality cannot be captured, and a new category may not be generated at the time of an abnormality. This becomes the cause of underreporting.

By using the model definition unit 400 included in the present invention, the normalization range can be appropriately determined in accordance with the operating conditions. This function suppresses false positives and improves diagnostic accuracy. Next, a specific method will be described.

FIG. 6 is a flowchart illustrating the operation of the operating condition determination unit 500 that is a component of the model definition unit 400 . As shown in FIG. 6 , the algorithm combines steps 510 , 520 , 530 , 540 , and 550 for execution.

First, in step 510, the data accumulated in the measurement signal 4 is divided into periods during constant load operation, during startup operation, during stop operation, and during load variation operation.

When the output does not change, that is, when the rate of change of the output measurement signal is small, it is assumed to be in constant load operation. In addition, it is divided into periods during start-up operation, stop operation, and load-change operation based on a signal for distinguishing start-up, stop, and load change included in the measurement signal of the power generation facility.

The process proceeds to step 520 during constant load operation, proceeds to step 530 during startup operation, proceeds to step 540 during stop operation, and proceeds to step 550 during load varying operation.

In step 520, information related to the load strap is extracted. For example, 0 to 50% of the rated output is set as low output, and 50 to 100% of rated output is set as high output, etc., and the load belts are classified according to output. In step 530, information on the type of start mode and the load change rate during start is extracted. The start mode includes a hot start mode, a cold start mode, and the like. In step 540, information related to the stop mode, the output when the stop operation starts, is extracted. As the stop mode, there are a mode in which the device is stopped during normal continuous use, a mode in which emergency shutdown is performed when an abnormality occurs, and the like. In step 550, information related to the load change rate, the output at the beginning of the load change, and the output at the end of the load change is extracted.

In addition, in this embodiment, in order to distinguish the states of constant load operation, start operation, stop operation, and load change operation in steps 520, 530, 540, and 550, the aforementioned information is extracted. Increase the amount of information. For example, as long as it is the information obtained by processing the measurement signal of the equipment 100 such as the turbine rotation speed, the acceleration rate, the type of fuel supplied to the equipment, and the atmospheric temperature, it may be added to the information obtained in steps 520, 530, 540, and 550. in the extracted information.

As shown in FIG. 6( b ), a diagnostic model 710 using the same measurement items is constructed from a collection of models classified into sub-diagnostic models 720 based on the information extracted in FIG. 6( a ). Different normalization conditions are defined for each sub-diagnosis model for diagnosis.

Next, an embodiment of the normalization condition determination unit 600 that determines the normalization condition will be described with reference to FIGS. 7 to 9 . The measurement signal is normalized by the data preprocessing device 710 using information on the normalization condition determined by the normalization condition determination unit 600 . Assume that the number of data items of the measurement signal xi is N, and the nth measurement signal is x(n). The normalized data Nxi(n) is represented by the following formula (13). Wherein, Nmin(n) is the lower limit value of normalization, and Nmax(n) is the upper limit value of normalization.

[number 13]

Nxi(n)=(xi(n)-Nmin(n))/(Nmax(n)-Nmin(n))...Formula (13)

In the normalization condition determination unit 600, Nmin(n) and Nmax(n) represented by the formula (13) are determined.

FIG. 7 is an explanatory diagram of a first embodiment of the normalization condition determination unit 600 .

FIG. 7( a ) is a flowchart of the first embodiment of the normalization condition determination unit 600 . As shown in Figure 7(a), this algorithm combines steps 611, 612, and 613 for execution.

First, in step 611 , the data for model construction is divided for each operating condition. In step 612, information on the variation range of the measurement signal is extracted for each operating condition.

In step 613, the upper limit Nmax1(n) of the normalization range and the lower limit Nmin1(n) of the normalization range are determined.

In the case of constant load operation, equations (14) and (15) are used. Here, Dmax1(n) is the maximum value of the measurement signal, and Dmin1(n) is the minimum value. α and β are constants.

[number 14]

Nmax1(n)=Dmax1(n)×(1+α)…Formula (14)

[number 15]

Nmin1(n)=Dmin1(n)×(1-β)…Formula (15)

When the load is changing, starting, or stopping, the output command value is matched, and the normalization range is determined as in equations (16) and (17).

[number 16]

Nmax2=MW*a+b...Formula (16)

[number 17]

Nmin2=MW*c+d...Formula (17)

Here, MW is the output command value, and a and c are the slopes when the measurement signal during the load change is first approximated. In addition, b and d are values calculated by equations (18) and (19).

[number 18]

b=f+Dmax2×1.2...Formula (18)

[number 19]

d=f-Dmin2×1.2...Formula (19)

Here, f is the intercept of the first approximation of the measurement signal during the load change, Dmax2 is the maximum value of the deviation between the straight line for the first approximation of the measurement signal and the measurement signal, and Dmin2 is the minimum value of the deviation.

FIG. 7( b ) is a diagram illustrating temporal changes in output and measurement signals when the load changes from low output to high output. The output is low until time T0a, the load changes from time T0a to time T0b, and the output is high after time T0b.

By operating the flow chart shown in Fig. 7(a), the upper limit value of the normalization range at low output is determined to be 1030, the lower limit value is determined to be 1040, and the upper limit value of the normalization range during load changes is The limit value was determined to be 1050, the lower limit value was determined to be 1060, the upper limit value of the normalization range at high output was determined to be 1070, and the lower limit value was determined to be 1080. In this way, the normalization range of the present invention changes in accordance with the operating conditions.

FIG. 8 is an explanatory diagram of the second embodiment of the normalization condition determination unit 600 . As shown in FIG. 8( a ), this algorithm is executed by combining steps 631 , 632 , and 633 .

In step 631, data of the load variation pattern is extracted. In step 632, dead time is estimated for each data item. The process signal is used as the input and output data, and the dead time is estimated from the impulse (impulse) response. As a specific calculation method, the method described in the reference document "Superior System Identification for Control" (Tokyo Denki University Publishing Bureau) ("控御のためのSuperior System Tongding" (Tokyo Denki University Publishing Bureau)) can be mentioned. . In step 633, the normalization range is determined for each data item after shifting the data by the unnecessary time estimated in step 632.

This state will be described using FIG. 8( b ). The timing at which the measurement signal starts to change is later than the timing T1a at which the output command value starts to increase. This time is meaningless time. The dead time of data item A is (T2a-T1a), and the dead time of data item B is (T3a-T1a).

After using the flow chart of FIG. 8( a ) to deviate the measurement signal by an amount of dead time, the flow chart of FIG. 7 is used to determine the normalization range. Accordingly, the normalization range of the measurement signal can be determined in accordance with the output command value.

FIG. 9 is an explanatory diagram of a third embodiment of the normalization condition determination unit 600 in the diagnostic device for power generation equipment shown in FIG. 1 .

The normalized range around the maximum value and the minimum value of the measurement signal is expanded, so that it is easy to detect the change of the measurement signal around the maximum value and the minimum value.

In addition, the normalization method of the normalization condition determination part 600 is not limited to the above-mentioned content, What is necessary is just to determine a normalization condition for every operating condition. For example, the range in which the data value changes is estimated from the design information of the equipment and the specification of the measuring device, and this range is used as the normalization range.

FIG. 10 is a flowchart illustrating an operation mode of the diagnostic device 200 for power generation equipment shown in FIG. 1 . FIG. 10( a ) is an operation flowchart of the model construction mode in FIG. 2 , and FIG. 10( b ) is an operation flowchart of the diagnosis mode.

As shown in FIG. 10( a ), the model building mode combines steps 1200 , 1210 , 1220 , and 1230 to execute. In step 1200 , data for a period used for model construction is extracted from the measurement signal database 310 . This period can be arbitrarily set by the operator of the facility 100 . Next, in step 1210, the operating condition determination unit 500 is operated. The flow chart shown in FIG. 6(a) operates to divide the data of the period used for model construction into each operation mode of constant load operation, start operation, stop operation, and load change operation, and further make the Steps 520, 530, 540, and 550 operate to divide the data of each pattern for each feature. Next, in step 1220, the normalization condition determination unit 600 is operated. The flow chart described in FIG. 7( a ) and FIG. 8( a ) is operated, and the normalization condition is determined for each group divided in step 1210 . The model definition information 5 obtained by operating steps 1210 and 1220 is stored in the model definition database 320 . Finally, in step 1230, the model building unit 700 is activated. Each group segmented in step 1210 is defined as a sub-diagnostic model 720 (see FIG. 6(b)), the measurement signal is processed using the normalization condition defined for each sub-diagnostic model, and the ART described in FIG. 3 is used to Build a diagnostic model.

As shown in FIG. 10( b ), the diagnostic mode combines steps 1300 , 1310 , and 1320 to execute. In step 1300 , data during a diagnosis period is extracted from the measurement signal database 310 . Next, in step 1310 , the data during the diagnosis is processed by the operating condition determination unit 500 . A sub-diagnostic model having the same operating condition is extracted from a plurality of sub-diagnostic models constructed in the model construction mode. Finally, in step 1320, the diagnosis unit 800 is activated. In the diagnosis unit 800 , the sub-diagnosis model information extracted in step 1330 is extracted from the model definition database 320 . The class numbers of the sub-diagnosis models are compared with the class numbers obtained by classifying the data during the diagnosis by ART. If it is the same category as when the model construction mode was activated, it will be diagnosed as normal, and if a different category number is generated, it will be diagnosed as abnormal. The diagnosis result 10 is output to the external output interface 220 .

Fig. 11 is a diagram illustrating the form of data stored in the database of the present invention.

As shown in FIG. 11(a), in the measurement signal database 310, the operation data obtained by measuring the equipment 100, that is, the measurement signal 1 (in the figure, described Values of data items A, B, C).

On the display screen 311, by using scroll bars 312 and 313 that can be moved vertically and horizontally, a wide range of data can be scrolled and displayed.

In the model definition database 320 , as shown in FIG. 11( b ), information on operating conditions and normalization ranges is stored in association.

In the diagnosis model database, as shown in FIG. 11( c ), the relationship between category numbers and weight coefficients is stored. Here, the weight coefficients are the center coordinates of the categories.

FIG. 12 is a diagram illustrating an application effect of the diagnostic device for power generation equipment shown in FIG. 1 .

The load change starts at time T4 and ends at time T5. After the load change ends, an abnormality occurs at time T6.

FIG. 12 graphically shows the relationship between the output command value and the turbine temperature, together with the relationship between the category number and the temperature measurement signal. The temperature rises with the occurrence of abnormality, but since it is within the range of category 4, it is classified as the same category as the normal state. In the conventional method, the normalization range is enlarged, and the temperature rise accompanying the occurrence of the abnormality cannot be captured, and the abnormality cannot be detected.

In the present invention, the operating conditions are matched to switch the diagnosis model. In this embodiment, the diagnosis is performed using the operating condition 1 diagnosis model (constant load operation, low output model) up to time T4, and the operation condition 2 diagnosis model is used between time T4 and T5 (load changing operation) Diagnosis is performed, and switching is performed to perform diagnosis using the operating condition 3 diagnosis model (constant load operation, high output model) after time T5.

In the operating condition 3 diagnosis model, normal states are classified into category numbers 1-5. Before the abnormality occurs, it is classified into the same category as the normal state, but after the abnormality occurs, the data is classified into a new category with category number 6, and the abnormality can be detected. That is, since the state is classified more finely than in the prior art, it is possible to suppress false negatives that cannot detect abnormalities.

FIG. 13 is a diagram illustrating a screen displayed on the image display device 940 in the diagnosis device 200 for power generation equipment shown in FIG. 1 .

Use the mouse 930 to operate the cursor 951, so that the upper limit value (954, 956, 958) of the normalized range, the lower limit value (955, 957, 959) of the normalized range, and the boundary time of the operating condition can be adjusted arbitrarily (952, 953). In order to reflect the adjustment result to the model, click the execute button 960 on the screen of FIG. 13 . Through this operation, the operating conditions and normalization range information of the model definition database shown in FIG. 10( b ) are changed, and the adjustment results can be reflected in the model construction and diagnosis operations.

In addition, this invention is not limited to the said Example, Various modification examples are included. For example, although the above-mentioned embodiments have been described in detail to explain the present invention easily, it is not limited to include all the configurations described.

In addition, each of the above-mentioned configurations, functions, processing units, processing units, etc. may be implemented in part or in whole by hardware such as integrated circuit design or the like. In addition, each of the configurations, functions, and the like described above can also be realized by software for a processor to interpret and execute a program that realizes each function. Information such as programs, tables, files, measurement signals, and calculation information for realizing each function can be stored in a storage device such as a memory or a hard disk, or a storage medium such as an IC card, SD card, or DVD. Therefore, each processing and each configuration can be realized as a processing component or a program module.

In addition, information lines are shown as deemed necessary for explanation, but not all control lines or information lines are necessarily shown on a product. In fact, it can also be considered that almost all the configurations are connected to each other.

According to this embodiment, it is possible to obtain a diagnostic device for a power generation facility and a method for diagnosing a facility that can detect abnormalities in the power generation facility with high precision.

Industrial availability

The present invention can be widely applied to various devices and the like as a device diagnosis device and a device diagnosis method.

Claims (14)

1. A diagnosis device for power generation equipment, which diagnoses the operating state of the equipment based on a measurement signal obtained by measuring the state quantity from the power generation equipment, and displays the diagnosis result on an image display device, wherein the diagnosis device for power generation equipment is characterized in that, have: a model constructing unit for constructing a model used in diagnosis using a measurement signal obtained by measuring a state quantity of the power generating equipment in a diagnostic device of the power generating equipment; a model definition unit that defines the operating conditions diagnosed by the model and the normalization method of the measurement signal; and a diagnosis unit for diagnosing the operating state of the power generation facility using the model constructed by the model construction unit, In the model definition unit have: an operating condition judging section that judges an operating condition of the power generating equipment; and a normalization condition determining unit that determines a normalization condition of the measurement signal for each operating condition determined by the operating condition determining unit, In the diagnosis unit, diagnosis is performed by switching the diagnosis model in accordance with the operating conditions. 2. The diagnostic device for power generation equipment according to claim 1, wherein: The operating condition judging unit has: An arithmetic device that classifies the operating conditions of the power generation equipment into any one of constant load operation, start operation, stop operation, and load variation operation; An arithmetic device that extracts the load band when the operating condition is constant load operation; An arithmetic device for extracting the type of starting mode and the rate of load change when the operating condition is that the starting operation is in progress; An arithmetic means for extracting a type of a stop mode and a stop operation start output while the stop is in operation; and An arithmetic unit that extracts the load change rate, the output at the start of the load change, and the output at the end of the load change while the load change is in operation. 3. The diagnostic device for power generation equipment according to claim 1, wherein: In the normalization condition determining unit, there are: a first computing device that divides the data for model construction for each operating condition determined by the operating condition determining unit; a second computing means for extracting information on the magnitude of data change for each operating condition; and A third calculation means for determining a normalization range based on the information of the data variation range. 4. The diagnostic device for power generation equipment according to claim 3, wherein: The diagnostic device of the power generating equipment includes a computing device that makes a normalization range of a measurement signal during a load change a function of an output command value. 5. The diagnostic device for power generation equipment according to claim 3, wherein: In the normalization condition determining unit, after operating an arithmetic means for estimating a dead time for a measurement signal during a load change, and an arithmetic means for shifting a measurement signal by the estimated dead time, the claim is made. The first computing device, the second computing device, and the third computing device described in 3 operate. 6. The diagnostic device for power generation equipment according to any one of claims 3 to 5, wherein: In the normalization condition determination unit, an arithmetic unit that expands a normalization range around a maximum value and a minimum value of the measurement signal is provided. 7. The diagnostic device for power generation equipment according to claim 1, wherein: The diagnostic device for power generation equipment includes an image display device for displaying a trend graph showing a temporal change of a measurement signal, a timing at which the operating condition determined by the operating condition determination unit is switched, and The normalization range determined by the normalization condition determining unit is superimposed and displayed, and is used to arbitrarily change the timing of switching the operating conditions and the normalization range. 8. A method for diagnosing power generating equipment, wherein the operating state of the equipment is diagnosed based on a measurement signal obtained by measuring a state quantity from the power generating equipment, and the diagnosis result is displayed on an image display device, the method for diagnosing the power generating equipment is characterized in that, With the following steps: an operating condition determination step for determining the operating conditions of the generating equipment; a normalization condition determination step of determining a normalization condition of the measurement signal for each operation condition determined by the operation condition determination unit; A step of constructing a model used in the diagnosis using a measurement signal obtained by measuring a state quantity of the power generation facility in a diagnosis device of the power generation facility; the step of defining the operating conditions for diagnosis by said model and the normalization method of the measured signals; and Diagnosis is performed by switching the diagnosis model according to the operating conditions. 9. The diagnostic method for power generation equipment according to claim 8, characterized in that: The step of determining the operating condition of the power generating equipment includes: a step of classifying the operating conditions of the power generating equipment into any one of constant load operation, start operation, stop operation, and load variation operation; The step of extracting the load belt under the condition that the operating condition is constant load operation; a step of extracting the type of starting mode and the load change rate when the operating condition is the starting operation; A step of extracting the kind of stop mode and stop action start output while the stop is in operation; and A procedure to extract the load change rate, the output at the start of the load change, and the output at the end of the load change while the load change is in operation. 10. The diagnostic method of power generation equipment according to claim 8, characterized in that, In the normalization condition determination step, have: a first step of dividing the data for model construction for each operating condition determined by the operating condition determining unit; a second step of extracting information on the magnitude of data variation for each operating condition; and The third step of determining the normalization range based on the information of the data variation range. 11. The method for diagnosing power generation equipment according to claim 10, wherein: The method includes a step of setting a normalized range of a measurement signal during a load change as a function of an output command value. 12. The method for diagnosing power generation equipment according to claim 10, wherein: In the normalization condition determination step, after executing the step of estimating dead time for the measurement signal in load variation and the step of shifting the measurement signal by the estimated dead time, the method described in claim 10 is executed. The first step, second step and third step. 13. The method for diagnosing power generation equipment according to any one of claims 10 to 12, wherein: In the normalization condition determination step, a step of expanding the normalization range around the maximum value and the minimum value of the measurement signal is included. 14. The method for diagnosing power generation equipment according to claim 8, wherein: It is possible to superimpose and display a trend graph showing the change over time of the measurement signal, the timing at which the operating condition determined by the operating condition determination unit is switched, and the normalization range determined by the normalization condition determination unit. The timing at which the operating conditions are switched and the normalization range are changed arbitrarily.