JPH04204198A - Signal selection for power plant monitoring system - Google Patents

Abstract

PURPOSE:To enable selection of a signal necessary for pursuit of cause by preserving a time-series data comprising a plurality of measuring signals from a power plant into a disc of a calculator to select a signal necessary for monitoring from on a cause-effect network between signals prepared. CONSTITUTION:A plurality of signals are latched through an A/D converter connected to a power plant 1 to detect a necessary abnormal condition by an abnormality detection means 3 and a data comprising a plurality of signals then obtained is preserved with a data preserver 4. Besides raw data values, a time-series data preserved is preserved together with a wave form pattern abbreviated, a name of signal, data collecting conditions, a name of abnormality detection trigger signal and the like concerning the data. Then, the selection of signals necessary for the pursuit of cause is performed with a data retrieving device 7 and the results are shown on a display 8. A network data base is prepared beforehand about a cause-effect network. This enables quick selection of a proper signal responding to conditions at emergency without overlooking it.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a signal selection method in a system for monitoring the behavior of a power plant using a computer.

(Prior Art) Various methods have been proposed for a power plant monitoring system using an online computer due to its large degree of freedom. Most of these monitoring systems measure measurement signals from power plants online in real time, perform some data processing, and then use appropriate monitoring judgment values to determine whether or not there is an abnormality. taking it. Further, when an abnormality is detected, data before and after the detection is stored in an external storage device and may be used to investigate the cause after the fact.

In this case, the problem is that the number of monitoring signals has increased from hundreds to thousands due to improvements in computer performance.
When an abnormality occurs, it is becoming difficult to select the signals necessary for investigating the cause and considering countermeasures after the fact.

(Problem to be Solved by the Invention) In most conventional monitoring systems, a certain number of signals are registered in advance as presets for each of the most prominent abnormal signals, and when an abnormality occurs, the preset group is displayed all at once. However, the situations in which abnormalities occur vary widely, and such a blizzet group alone cannot provide the signals necessary to ultimately investigate the cause. There are often times when things get overlooked.

In a power generation plant, it is important to quickly and reliably detect abnormalities at the time of their initial signs, take immediate countermeasures, and prevent the abnormalities from developing into larger abnormalities. In such cases, we can select all the signals necessary to investigate the cause within a limited amount of time.
For use in monitoring, a flexible signal selection method that adapts to the state of abnormal symptoms is required. In particular, by carrying out all the necessary monitoring once a small abnormality is detected, it is possible to prevent it from developing into a larger problem.

Under these circumstances, power plant monitoring equipment can save the behavior of plant monitoring signals when an abnormality is detected.
Furthermore, there is a strong demand for a method that can quickly select signals whose causes need to be investigated from among a large number of signals without overlooking them.

The present invention has been made in view of the above points, and its purpose is to quickly and completely select signals necessary for investigating the cause from data consisting of a large number of signals detected and stored by a computer. An object of the present invention is to provide a signal selection method in a power plant monitoring system.

[Structure of the Invention] (Means for Solving the Problem) A signal selection method in a power generation plant monitoring system stores time-series data consisting of a plurality of measurement signals from a power generation plant on a computer disk, and selects a signal from the stored data. ,
Signals necessary for monitoring are selected from a causal network between signals prepared in advance.

(Function) First, as shown in FIG. 1, the computer takes in a plurality of signals through the A/D converter (analog-digital converter) 2 connected to the power generation plant 1 online.
A necessary abnormal state is detected by some abnormality detection means 3, and data consisting of a plurality of signals at that time is stored in a data storage device 4, which is an external storage medium such as a magnetic storage device. In some cases, data stored by another computer system 5 may be transferred and stored through a communication management device 6 such as a local network. In addition to raw data values, the time-series data thus saved is saved together with its reduced waveform pattern, signal name, data collection condition, abnormality detection trigger signal name, etc. Figure 4 shows the structure of this data. Next, the data retrieval device 7 selects signals necessary for investigating the cause and displays them on the display device 8.

The following four selection methods are available for signal name selection in the data search device 7.

(1) Select from preset groups prepared in advance.

(2) Select from the signal names displayed on the plant mimic diagram (block diagram).

(3) Prepare a list table of signal names and select from it.

(4) Display the causal relationship network of the attention signal and select from it.

The concept of this selection method is shown in FIGS. 2 (1) to (4).

In addition to displaying the signal name, both diagrams also display the reduced pattern of the signal, which allows you to intuitively understand the behavior of the signal and use it to make selection decisions. can. In particular, regarding the causal relationship network which is the focus of the present invention, a network database as shown in FIG. 3 is prepared in advance. In addition, the input-output causality of signals is not clear, and signals that have a similar relationship such as signals of the same type (
A network of related signals) is also available, allowing you to select the necessary signal from among multiple relationships.

(Example) Embodiments of the present invention will be described below with reference to the drawings.

First, the configuration of an embodiment of the present invention is shown in FIG.

A plurality of signals obtained from the power generation plant 1 are transmitted from an A/D converter 2 connected to the power generation plant 1 online.
Alternatively, data stored by another computer system 5 is passed through a communication management device 6, which is a communication means such as a local area network, and then stored in the data storage device 4, which is a magnetic storage device. As shown in Figure 4, the time series data saved in this way includes, in addition to the raw data,
Its reduced waveform pattern, signal name, data collection conditions,
It is saved along with the abnormality detection trigger signal name, etc.

Here, the contracted pattern is displayed in the form shown in FIG. 5, and the method for obtaining it is as follows.

Specifically, N time-series data of the signal of interest is expressed as (X(x
), t-1, N), its maximum and minimum values are XMIN, N
(7) X (t) (7) Minimum value XMAX, Let be the maximum value of N X(t). At this time, when N pieces of data are divided into M pieces, each block will contain BN-N/M pieces of data. The minimum and maximum values of each block are CMIN(1)=(X(j),j-(i-t*BN)
, i*BN)) (however, i-1, M) CNAX(j)-(X(j), j-(j-1*BN,
i*BN)) (however, 1-x, M). Set this value to MIN using an appropriate integer value MM.
(i)=(CMIN(i)-XMIN)/(XMAX
-XMIN)*MM (round down to the nearest whole number) MAX(i)-(CMAX(i)-XMIN)/(X
MAX-XMIN)*MM (round down to the nearest whole number) (where i=1, M), 2*M sets of integer values in the range 0-MM are obtained. Therefore, it is reduced to N data strings (usually on the order of several tens) or 2*M (usually on the order of several tens). In this display, as shown in FIG. 5, M blocks are plotted on the horizontal axis, and the minimum value to maximum value are filled in on the vertical axis for MM divisions. Maximum total data
Since it is normalized by the minimum value, as shown in Figure 5,
In the case of steady data, the entire area is filled in, and in the case of transient data, there are many white areas. In other words, since appropriate scaling is performed according to the data situation, it is possible to appropriately understand the waveform pattern with a small amount of data. In addition, since patterns can be displayed in a very short time, even if they are used bare-handed with the display of signal names, there is no problem with interaction with the user.

Next, the data search device 7 selects the signals necessary for investigating the cause.
This is done using the following four selection methods.

(1) Select from preset groups prepared in advance.

(2) Select from the signal names displayed on the plant mimic diagram (block diagram).

(3) Prepare a list table of signal names and select from it.

(4) Display the causal relationship network of the attention signal and select from it.

The concept of this selection method is shown in FIGS. 2 (1) to (4).

In the preset method (1), a comment indicating the meaning is attached to each preset, and when one of the comments is selected, a reduced waveform pattern of the signal name included in the preset is displayed. By looking at this signal name and waveform pattern, it is determined whether the selected group is appropriate.

(2) Signals can be selected from the signal names and waveform patterns displayed on the block diagram while intuitively grasping the meanings of the signals, physical causal relationships, and waveform patterns. However, for this purpose, appropriate signal selection may not be possible unless someone understands the physical meaning of the block diagram and the causal relationships between signals. In particular, it is thought that inexperienced users are more likely to overlook selections, and when making selections across multiple block diagrams, the possibility of oversights is particularly high.

(3) Selection from the signal name list table is the most primitive method, and is fine if you know which signal to select, but if not, you have a sufficient grasp of the causal relationships between the signals. Moreover, it requires knowing which signals are included in the data and selecting them, which is the most labor-intensive and
Moreover, it is a difficult method that requires a skilled specialist.

Method (4) is a method to compensate for the shortcomings of the above methods.
This displays the causal relationship network of the signal of interest and selects from it. First, the signal of interest is selected using the methods (2) and (3) above. Also, a bird in the database may use a single signal. Next, in order to investigate the cause, it is necessary to appropriately select other signals related to this signal of interest. For this purpose, the causal relationship network database between signals shown in FIG. 3 is used.

FIG. 3 shows two types of relationships. One is the physical causal relationship (input/output relationship) between signals, and the other is related similar signals. This corresponds to, for example, multiplexed signals of the same type. In the case of input/output signals, tracing the signals one after another from the input side to the output side results in a tree-like display. On the other hand, similar signals constitute one group for each signal of interest.

This situation is shown in FIG. 2 (4). With such a display, it is possible to select an appropriate signal according to the abnormal situation by considering the physical causal relationship between the signals and displaying the reduced waveform pattern at the same time. Also, the third
The causal relationship network shown in the figure does not need to be created for each power generation plant to be monitored, because once the type of diagnosis target is determined, the data will be determined exactly as expected. If the monitoring signal is different for each monitoring power plant, when displaying in Figure 2 (4), AND the monitoring signal of each power plant and the signal on the causal relationship network in Figure 3, and if the If there is no monitoring signal on the related network, the corresponding signal can also be searched for by further tracing the network.

Therefore, even if the monitoring signals are different for each power plant, the same algorithm can be used.

With the recent improvement in computer performance, the number of monitoring signals in online monitoring systems for power plants is increasing, and the number of monitoring signals is on the order of several hundred to several thousand. Furthermore, the types of signals are not necessarily common among monitoring systems of the same type, and are combined in some specific ways depending on the purpose of monitoring. On the other hand, the number of signals that humans can see at the same time is at most 10 to 20 signals. Therefore,
It is necessary to select 10 to 20 signals for online monitoring and post-event monitoring from the huge number mentioned above, but the required signals vary depending on the abnormality situation, and Since the types of signals being monitored differ depending on the monitoring system, it can be difficult even for experienced professionals to select the necessary signals without overlooking them.

Conventionally, in order to deal with this situation, a certain number of signals are often allocated in advance in the form of a preset group for each monitoring purpose, and the necessary one is selected from the preset group. In this way,
Required signals may be overlooked. In the past, to compensate for this, experts involved in the design and operation of power plants or their experiential knowledge were required to manually
Selecting the necessary signal. The specific expert knowledge used includes physical causal relationships between monitoring signals, related signals experienced in past troubles, and the names of signals actually input to the target monitoring system. It is thought that

In light of the above circumstances, the method of the present invention, which supports the appropriate selection of signals necessary for plant diagnosis, improves monitoring efficiency,
This will be extremely effective from the viewpoint of speeding up the process and preventing oversights.

[Effects of the Invention] As explained above, the present invention provides a time-series data storage method, particularly a reduced pattern storage method, a signal selection method using the display, and a signal selection method using a signal causality network. By selection, it is possible to quickly select an appropriate signal corresponding to the situation at the time of an abnormality without overlooking it. Furthermore, even inexperienced users can select appropriate signals by using the causal relationship network.

Appropriate signal selection without oversight will lead to prompt and careful data review when abnormal symptoms occur, and by investigating the cause of the abnormality, it will be of great use in improving plant availability and ensuring reliability.

[Brief explanation of the drawing]

Fig. 1 shows the system configuration of the present invention, Fig. 2 shows four methods of signal selection, Fig. 3 shows the database structure of the signal causality network for signal selection, and Fig. 4 shows the system configuration of the present invention. The figure shows a time-series database structure in the event of an abnormality, and FIG. 5 shows a method of displaying signal reduction patterns used in signal selection. 1... Power plant 2... A/D converter 3... Abnormality detection means 4... Data storage device 5...・Computer system 6...Communication management device 7...Data search device 8...Display device

Claims (1)

[Claims] (1) Time-series data consisting of multiple measurement signals from a power generation plant is saved on a computer disk, and from the saved data, signals necessary for monitoring are selected from a causal network between signals prepared in advance. A signal selection method in a power generation plant monitoring system characterized by: