Specification
Title of Invention: TIME SERIES DATA PROCESSING DEVICE AND METHOD
THEREFOR
Technical Field
[0001]
The present invention relates to a time series data processing unit and
method for processing time series data generated by, for example, various sensors.
Background Art
[0002]
Conventionally, with computer systems and the like, data detected by
various sensors are fetched as time series data and, for example, analysis and
control processing is executed by using the fetched time series data. For example, a
technique that collects and analyzes a plurality of pieces of time series data, which
are fetched at various places, is suggested (see Patent Literature 1 ).
[0003]
Patent Literature 1 discloses a technique used when analyzing a plurality of
pieces of time series data to: complement missing data according to sampling
cycles of the plurality of pieces of time series data in consideration of the fact that
each piece of data has a different sampling cycle; and accumulate the
complemented data. Specifically speaking, if the time series data are sampled in a
1-second cycle or a 2-second cycle, the time series data in the 2-second cycle are
complemented based on the time series data in the 1-second cycle, thereby
accumulating each piece of time series data.
[0004]
Furthermore, when efficiently displaying long-term plant data at high speeds,
a technique that efficiently searches a large amount of long-term data generated at
a plant, monitors tendencies of the plant data, and monitors abnormal values is
suggested (see Patent Literature 2).
[0005]
Patent Literature 2 discloses a technique that fetches the plant data in a
2
specified data fetching cycle, writes them to a plant data table, fetches the plant data
in a data record cycle longer than the data fetching cycle, collects them in each data
record cycle, stores them in a plant data history information table, creates long-term
search history information including a maximum value, minimum value, and average
value of the plant data for each piece of plant data history information, stores it in a
long-term search history information table, performs a data search of any one of the
tables when displaying the data, and displays the content of the searched data as a
graph on a display device.
Citation List
Patent Literature
[0006]
Patent Literature 1: Japanese Patent Application Laid-Open (Kokai) Publication No.
2003-44518
Patent Literature 2: Japanese Patent Application Laid-Open (Kokai) Publication No.
2010-49533
Disclosure of Invention
Problems to be Solved by the Invention
[0007]
Regarding Patent Literature 1, the time series data are complemented in
accordance with a shorter sampling cycle. So, even if each piece of data has a
different sampling cycle, the time series data can be analyzed accurately. However,
since the time series data are accumulated in accordance with a shorter cycle upon
accumulation of the time series data, an amount of accumulated data increases and
it becomes difficult to efficiently accumulate the plurality of pieces of time series
data.
[0008]
Furthermore, the technique of Patent Literature 2 calculates the average
value, the maximum value, or the minimum value of each piece of time series data
in each constant cycle. So, if it is necessary to perform analysis by using the plurality
of pieces of time series data, for example, if operations to calculate differences
3
between the plurality of pieces of time series data (such as a difference between the
maximum value and the minimum value) are required, it is necessary to access
individual pieces of the time series data and calculate the values of each piece of
time series data. Therefore, the plurality of pieces of time series data cannot be
accessed efficiently.
[0009]
The present invention was devised in light of the above-described problems
of the conventional technology and it is an object of the invention to provide a time
series data processing unit and method capable of efficiently accessing the plurality
of pieces of time series data which are analysis targets.
Means for Solving the Problems
[0010]
In order to achieve the above-described object, the present invention is
characterized in that: upon data accumulation, a plurality of pieces of time series
data which are analysis targets are combined, the combined plural pieces of time
series data are accumulated in a storage unit by associating them with their
attribute; and upon data analysis, the combined plural pieces of time series data are
extracted, as time series data to be used for the analysis, based on the attributes
from the storage unit.
Advantageous Effects of Invention
[0011]
According to the present invention, the plurality of pieces of time series data
to be used for analysis can be accessed and analyzed efficiently.
Brief Description of Drawings
[0012]
[Fig. 1]
Fig. 1 is a block diagram showing the overall configuration of a computer
system to which the present invention is applied.
[Fig. 2]
Fig. 2 is a structure diagram of a time series data management table.
4
[Fig. 3]
Fig. 3 is a structure diagram of attribute-aggregated information.
[Fig. 4]
Fig. 4 is a structure diagram of an attribute buffer.
[Fig. 5]
Fig. 5 is a structure diagram of time-aggregated information.
[Fig. 6]
Fig. 6 is a structure diagram of a time buffer.
[Fig. 7]
Fig. 7 is a structure diagram of an aggregated data table.
[Fig. 8]
Fig. 8 is a structure diagram of an analysis query.
[Fig. 9]
Fig. 9 is a flowchart for explaining time series data accumulation processing.
[Fig. 10]
Fig. 10 is a flowchart for explaining time series data analysis processing.
[Fig. 11]
Fig. 11 is a flowchart for explaining processing for creating lists of an
analysis target ID, a search target ID, an acquisition target ID, and an acquisition
target aggregation I D.
[Fig. 12]
Fig. 12 is a flowchart for explaining data acquisition processing.
[Fig. 13]
Fig. 13 is a flowchart for explaining data extraction and analysis processing
according to a condition and an attribute.
[Fig. 14]
Fig. 14 is a block diagram showing an overall configuration diagram of a
computer system according to a second embodiment.
[Fig. 15]
Fig. 15 is a structure diagram of cycle-aggregated information.
5
0 [Fig. 16]
Fig. 16 is a structure diagram of a cycle buffer.
[Fig. 17]
Fig. 17 is a structure diagram of attribute-aggregated information.
[Fig. 18]
Fig. 18 is a structure diagram of an attribute buffer.
[Fig. 19]
Fig. 19 is a structure diagram of time-aggregated information.
[Fig. 20]
Fig. 20 is a structure diagram of a time buffer.
[Fig. 21]
Fig. 21 is a structure diagram of an aggregated data table.
[Fig. 22]
Fig. 22 is a structure diagram of an analysis query.
[Fig. 23]
Fig. 23 is a flowchart for explaining time series data accumulation
processing.
[Fig. 24]
Fig. 24 is a flowchart for explaining time series data analysis processing.
[Fig. 25]
Fig. 25 is a flowchart for explaining data extraction and analysis processing
according to a condition, an attribute, and a cycle.
Description of Embodiments
Embodiment 1
[0013]
This embodiment is designed so that: upon data accumulation,
combinations of time series data generated in the same cycle (time series data
relating to a predicted number of vibrations and an actual number of vibrations) are
selected as combinations of a plurality of pieces of time series data, which become
analysis targets, a plurality of sets of the selected combinations of the time series
6
data are aggregated on a set time basis, and the plurality of sets of the aggregated
time series data are accumulated in a storage unit by associating with an attribute
(the number of vibrations); and upon data analysis, the combinations of time series
data generated in the same cycle (time series data relating to the predicted number
of vibrations and the actual number of vibrations) are extracted, as time series data
to be used for the analysis, from the storage unit.
[0014]
An embodiment of the present invention will be explained below based on
drawings.
[0015]
Fig. 1 is a block configuration diagram of a computer system to which the
present invention is applied. Referring to Fig. 1, the computer system is configured
by including a time series data source 10, a client computer 12, a network 14, a time
series data processing unit 16, and an external storage device 18. The time series
data source 1 0, the client computer 12, and the time series data processing unit 16
are connected to each other via the network 14; and the time series data processing
unit 16 is connected to the external storage device 18.
[0016]
The time series data source 10 includes, for example, various sensors such
as sensors for detecting temperatures, humidity, voltages, electric currents, power
generation, power consumption, and an actual number of turbine vibrations, or a
predicted-number-of-vibrations generating vibration machine for generating, for
example, a predicted value of turbine vibrations; and is configured as a time series
data generation source that outputs output signals from the various sensors and the
predicted-number-of-vibrations generating vibration machine as time series data to
the network 14 in chronological order.
[0017]
The client computer 12 has, for example, a processor, a memory, an
input/output device, a storage device, and a display device; issues an analysis query
as an analysis request to the network 14 and receives data from the time series data
7
processing unit 16 via the network 14, and stores the received data as result data in
the storage device.
[0018]
The time series data processing unit 16 is constituted from a memory 20, a
communications interface 22, an external storage interface 24, and a processor 26;
the memory 20, the communications interface 22, the external storage interface 24,
and the processor 26 are connected to each other via an internal network 28; the
communicatiqns interface 22 is connected to the network 14; and the external
storage interface 24 as a storage unit is connected to the external storage device 18.
Incidentally, the following configuration can be employed so that a storage device as
a storage unit instead of the external storage device 18 may be located in the time
series data processing unit 16 and this storage device may be connected to the
internal network 28.
[0019]
The processor 26 supervises and controls the entire time series data
processing unit 16 and executes various processing in accordance with a time
series data processing program 30 stored in the memory 20.
[0020]
Under this circumstance, the processor 26 functions as a data processing
unit for sequentially inputting and processing time series data, which are output from
the time series data source 1 0; and also functions as a data acquisition analysis unit
for accessing an aggregated data table 32 stored in the external storage device 18,
obtaining data from the aggregated data table 32, and analyzing the obtained data.
[0021]
The time series data processing program 30 is constituted from a data
accumulation unit 34, a data analysis unit 36, a data acquisition unit 38, and a
setting information storage area 40.
[0022]
The data accumulation unit 34 functions as a data processing unit and is
constituted from a data reception unit 42, a data attribute aggregation unit 44, a data
8
time aggregation unit 46, a characteristic point extraction unit 48, a data
compression unit 50, an aggregated data write unit 52, an attribute buffer 54, and a
time buffer 56.
[0023]
The data analysis unit 36 is composed of an analysis reception unit 60 and
an analysis execution unit 62.
[0024]
The data acquisition unit 38 is constituted from a read time zone
narrowing-down unit 70, an aggregated data read unit 72, a data unzipping unit 7 4,
a data time extraction unit 76, a data narrowing-down unit 78, and a data attribute
extraction unit 80. Under this circumstance, the data analysis unit 36 and the data
acquisition unit 38 function as a data acquisition analysis unit.
[0025]
The setting information storage area 40 stores attribute-aggregated
information 90 and time-aggregated information 92.
[0026]
If the time series data which are output from the time series data source 1 0
are input to the time series data processing unit 16 via the network 14 and the
communications interface 22, the data reception unit 42 for the data accumulation
unit 34 sequentially accepts the input time series data in chronological order. When
receiving time series data relating to a predicted number of vibrations and time
series data relating to an actual number of vibrations as time series data generated
in the same cycle, this data reception unit 42 outputs each pieces of the time series
data to the data attribute aggregation unit 44.
[0027]
The data attribute aggregation unit 44: combines a plurality of pieces of time
series data, which become analysis targets, that is, a plurality of pieces of time
series data generated in the same cycle, among the time series data which were
input; combines the combined plural pieces of time series data, for example, the
time series data relating to the predicted number of vibrations and the actual number
9
of vibrations; aggregates the combined plural pieces of time series data by
associating them with a mutually related attribute, for example, the number of
vibrations (hereinafter sometimes referred to as attribute-aggregated); and
accumulates the aggregated time series data as attribute-aggregated data in the
attribute buffer 54 and outputs them to the data time aggregation unit 46.
[0028]
The data time aggregation unit 46 gathers and collects the
attribute-aggregated data, which have been input, only for a set time period (set
time) and processes them as time-aggregated data. For example, the data time
aggregation unit 46 aggregates the attribute-aggregated data, which have been
input, on an hourly basis, accumulates them as time-aggregated data in the time
buffer 56, and outputs them to the characteristic point extraction unit 48.
[0029]
Specifically speaking, if the time series data, which are the analysis targets
and are a plurality of pieces of time series data generated in the same cycle, are
combined, the data attribute aggregation unit 44 selects one set of the combinations
of the time series data each time, gathers and aggregates each selected set of the
time series data on an hourly basis, which is the set time, and accumulates the
aggregated plural sets of combinations of the time series data as the
time-aggregated data in the time buffer 56.
[0030]
The characteristic point extraction unit 48 extracts characteristic points from
the time-aggregated data, which have been input, and outputs the extracted
characteristic points such as a maximum predicted value, a minimum predicted
value, a maximum actual value, and a minimum actual value as well as the input
time-aggregated data to the data compression unit 50.
[0031]
The data compression unit 50 compresses the time-aggregated data, which
have been output from the characteristic point extraction unit 48, and outputs the
compressed data and the characteristic points, which have been output from the
10
characteristic point extraction unit 48, to the aggregated data write unit 52.
[0032]
The aggregated data write unit 52 writes the data, which have been output
from the data compression unit 50, as aggregated data to the aggregated data table
32 of the external storage device 18.
[0033]
When an analysis query issued from the client computer 12 is input via the
network 14 and the communications interface 22, the analysis reception unit 60 for
the data analysis unit 36 receives the analysis query from the client computer 12
and outputs the received analysis query to the data acquisition unit 38. The analysis
execution unit 62 executes analysis as specified by the analysis query based on the
processing results of the data acquisition unit 38.
[0034]
The read time zone narrowing-down unit 70 for the data acquisition unit 38
interprets the analysis query from the analysis reception unit 60, obtains a read time
zone and search conditions specified by the analysis query, narrows down the read
time zone based on the characteristic points of the read time zone obtained from the
aggregated data table 32, and outputs it to the aggregated data read unit 72.
[0035]
The aggregated data read unit 72 searches the aggregated data table 32
based on the read time zone narrowed down by the read time zone narrowing-down
unit 70 and outputs data belonging to a data aggregation time zone stored in the
aggregated data table 32 (data including, for example, the aggregation ID, the
characteristic points, and the time-aggregated data) to the data unzipping unit 74.
[0036]
The data unzipping unit 74 unzips the time-aggregated data output from the
aggregated data read unit 72 and outputs the unzipped time-aggregated data to the
data time extraction unit 76.
[0037]
The data time extraction unit 76 extracts time-aggregated data of a read
11
time zone from the time-aggregated data unzipped by the data unzipping unit 7 4,
processes the extracted time-aggregated data as attribute-aggregated data, and
outputs the attribute-aggregated data to the data narrowing-down unit 78.
[0038]
The data narrowing-down unit 78 narrows down and extracts the
attribute-aggregated data, which satisfy the search conditions specified by the
analysis query, from the attribute-aggregated data processed by the data time
extraction unit 76, and outputs the narrowed-down attribute-aggregated data to the
data attribute extraction unit 80.
[0039]
The data attribute extraction unit 80 extracts a plurality of pieces of time
series data (time series data relating to the predicted number of vibrations and the
actual number of vibrations), which have the analysis target attribute (the number of
vibrations), for each same cycle from the attribute-aggregated data narrowed down
by the data narrowing-down unit 78 and outputs the plural pieces of the extracted
time series data to the analysis execution unit 62.
[0040]
Next, Fig. 2 shows the structure of a time series data management table.
[0041]
Referring to Fig. 2, a time series data management table 100 is a table
stored in, for example, the memory 20 for the time series data processing unit 16
and is constituted from a name 102, an attribute 104, a date and time 106, and a
value 108. This time series data management table 100 sequentially stores
information about time series data generated from the time series data source 10.
[0042]
For example, time series data 01 is managed as data regarding which the
name 102 is a "predicted value of turbine vibrations," the attribute 104 is a "predicted
number of vibrations," the date and time 106 is "2010-05-01 07:00:00," and the
value 108 is "15.2". Time series data 02 is managed as data regarding which the
name 102 is an "actual value of turbine vibrations," the attribute 104 is an "actual
12
number of vibrations," the date and time 106 is "2010-05-01 07:00:00," and the
value 108 is "24.3".
(0043]
Time series data 03 is managed as data regarding which the name 102 is a
"watt checker," the attribute 104 is "power consumption," the date and time 106 is
"201 0-05-01 07:00:00," and the value 108 is "6. 7"; and time series data 04 is
managed as data regarding which the name 1 04 is a "power generation measuring
instrument," the attribute 104 is "power generation," the date and time 106 is
"201 0-05-01 07:00:00," and the value 108 is "240."
[0044]
The time series data 01, 02 are time series data generated in the same
cycle; and in the next generation cycle, they will be managed as time series data 05,
06.
[0045)
The time series data 03, 04 are time series data generated in mutually
different cycles. The time series data 03 is generated every minute; and in the next
generation cycle, it will be managed as time series data 07. The time series data 04
is time series data generated in a one-hour cycle; and in the next generation cycle, it
will be managed as time series data 09.
[0046]
Next, Fig. 3 shows the structure of attribute-aggregated information 90
stored in the setting information storage area 40.
[0047]
Referring to Fig. 3, the attribute-aggregated information 90 is constituted
from an aggregation 10 120 and an attribute 122. The attribute 122 stores, for
example, the "predicted number of vibrations" as an attribute of time series data
relating to the predicted value of turbine vibrations and the "actual number of
vibrations" indicating the attribute of the actual value of turbine vibrations among the
plurality of pieces of time series data which become the analysis targets. The
aggregation 10 120 stores, for example, the "number of vibrations" which is an
13
attribute indicated by, and mutually related to, the predicted number of vibrations
and the actual number of vibrations.
[0048]
Next, Fig. 4 shows the structure of the attribute buffer 54.
[0049]
An information storage area of the attribute buffer 54 is configured in a table
format and the table is constituted from a date and time field 130, an attribute field
132, and a value field 134. Each entry of the date and time field 130 stores
information about the date and time when a plurality of pieces of time series data
which become the analysis targets are obtained.
[0050]
Each entry of the attribute field 132 stores, for example, the "predicted
number of vibrations" and the "actual number of vibrations" as information about the
attribute of the plurality of pieces of time series data which become the analysis
targets.
[0051]
Each entry of the value field 134 stores information about the value of each
piece of time series data, for example, "15.2" as the value of the time series data
relating to the predicted number of vibrations and "24.3" as the value of the time
series data relating to the actual number of vibrations.
[0052]
Next, Fig. 5 shows the structure of the time-aggregated information 92
stored in the setting information storage area 40.
[0053]
The time-aggregated information 92 is constituted from an aggregation ID
140 and the number of time-aggregated data pieces 42. The aggregation ID 140
stores, for example, the "number of vibrations" which is an attribute indicated by, and
mutually related to, the predicted number of vibrations and the actual number of
vibrations. The number of time-aggregated data pieces 142 stores, for example,
information indicating "3600 pieces" as the number of pieces of time series data to
14
be accumulated for one hour.
[0054]
Next, Fig. 6 shows the structure of the time buffer 56.
[0055]
An information storage area of the time buffer 56 is configured in a table
format and the table is constituted from an aggregation 10 field 150, a data
aggregation time zone field 152, and an attribute-aggregated data field 154.
[0056]
An entry of the aggregation 10 field 150 stores, for example, information of
the "number of vibrations." An entry of the data aggregation time zone field 152
stores information about a time zone during which the time series data are
aggregated.
[0057]
Each entry of the attribute-aggregated data field 154 stores a plurality of
pieces of time series data which become the analysis targets, for example, a
combination of the time series data 01 and the time series data 02 by associating
the combined time series data 01, 02 with the "number of vibrations" which is the
aggregation 10.
[0058]
Next, Fig. 7 shows the structure of the aggregated data table 32 stored in
the external storage device 18.
[0059]
Referring to Fig. 7, the aggregated data table 32 is constituted from an
aggregation 10 field 160, a characteristic point field 162, a data aggregation time
zone field 164, and a time-aggregated data field 166.
[0060]
For example, each entry of the aggregation 10 field 160 stores information
of the "number of vibrations" which is an attribute aggregating the attribute indicated
by the predicted number of vibrations and the actual number of vibrations.
[0061]
15
Each entry of the characteristic point field 162 stores information about
characteristic points of the time series data accumulated for one hour among the
plurality of pieces of time series data which have been input. For example, a
maximum value of the predicted number of vibrations is stored as a "maximum
predicted value: 30"; a minimum value of the predicted number of vibrations is
stored as a "minimum predicted value: 0"; a maximum value of the actual number of
vibrations is stored as a "maximum actual value: 40"; and a minimum value of the
actual number of vibrations is stored as a "minimum actual value: 0".
[0062]
The data aggregation time zone field 164 stores information about a data
aggregation time zone, during which each piece of time series data was gathered as
the time-aggregated data, as a numerical value together with information indicating
the relevant date.
[0063]
Each entry of the time-aggregated data field 166 stores data relating to the
time-aggregated data which were aggregated on an hourly basis. Under this
circumstance, the time series data 01, 02 are combined, the time series data 05,
06 are combined, and the respectively combined time series data 01, 02, 05, 06
are stored by associating them with the "number of vibrations" which is the
aggregation 10. Furthermore, each entry stores the combined time series data for
one hour.
[0064]
Next, Fig. 8 shows the structure regarding an analysis query.
[0065]
Referring to Fig. 8, an analysis query 170 is constituted from a selected
range (select_range) 172, selected items (select_items) 174, a data acquisition
target time zone (from_timerange) 176, and a search condition (where_condition)
178.
[0066]
The selected range 172 stores, for example, "1 second" as timing of
16
processing by the data analysis unit 36. The selected items 17 4 store, for example,
"Predicted Frequency - Actual Frequency AS Divergence Frequency" as the
analysis target ID for the analysis execution unit 62 to perform analysis.
[0067]
The data acquisition target time zone 176 stores, for example, "201 0-05-01
07:20:00 to 2010-05-01 08:40:00" as information about the data acquisition target
time zone used by the analysis execution unit 62 when executing the analysis.
[0068]
The search condition 178 stores information of, for example, the "predicted
number of vibrations>= 40" as a condition to become the search target I D.
[0069]
Next, the time series data accumulation processing will be explained in
accordance with a flowchart of Fig. 9.
[0070]
This processing is started by activation of the data accumulation unit 34 of
the time series data processing program 30 by the processor 26.
[0071 1
Firstly, when the data reception unit 42 receives time series data from the
time series data source 10 via the communications interface 22, the data reception
unit 42 sequentially delivers the received time series data to the data attribute
aggregation unit 44 (S 11 ).
[0072]
Next, the data attribute aggregation unit 44 compares the attribute of a
plurality of pieces of time series data in the same cycle among the received time
series data, for example, "the predicted number of vibrations and the actual number
of vibrations" with the attribute 122 of the attribute-aggregated information 90; and if
the attribute of each piece of the received time series data is the "predicted number
of vibrations" or the "actual number of vibrations," the data attribute aggregation unit
44 obtains the "number of vibrations" as the aggregation ID corresponding to the
attribute of each piece of the received time series data (812) and accumulates each
17
piece of the received time series data in the attribute buffer 54 corresponding to the
aggregation ID (the number of vibrations) (513).
[0073]
Next, the data attribute aggregation unit 44 judges whether all attribute
values on the same date and time exist in the attribute buffer 54 corresponding to
the aggregation ID or not (514); and if all the attribute values do not exist, the data
attribute aggregation unit 44 returns to the processing in step 511; and if it is
determined that all the attribute values exist, the data attribute aggregation unit 44
obtains data of all the attributes on the same date and time from the attribute buffer
54 corresponding to the aggregation ID and then deletes the data in the attribute
buffer 54 (515).
[0074]
Next, the data attribute aggregation unit 44: performs attribute aggregation
to combines the data of all the attributes obtained from the attribute buffer 54 for
each of the plural pieces of time series data which become the analysis targets and
associate them with the aggregation ID; processes the combined plural pieces of
the time series data as attribute-aggregated data; and outputs these
attribute-aggregated data to the data time aggregation unit 46 (516).
[0075]
Next, the data time aggregation unit 46 receives the attribute-aggregated
data and accumulates these attribute-aggregated data in the time buffer 56 (517),
and judges whether or not the number of data pieces of the time buffer 56 exceeds
the number of pieces stored in the number of time-aggregated data pieces 42 of the
time-aggregated information 92, for example, 3600 pieces (518); and if the number
of data pieces of the time buffer 56 does not exceed the number of time-aggregated
data pieces, the data time aggregation unit 46 returns to the processing in step 511
in order to collect data for one hour; and if the number of data pieces of the time
buffer 56 exceeds the number of time-aggregated data pieces, the data time
aggregation unit 46 proceeds to processing in step 519, recognizing that the data
for one hour has been collected.
18
[0076]
Next, in step 519, the data time aggregation unit 46 obtains all pieces of the
attribute-aggregated data from the time buffer 56 corresponding to the aggregation
10 (the number of vibrations) and then deletes the data in the time buffer 56.
[0077]
Next, the data time aggregation unit 46 aggregates time of all the pieces of
the collected attribute-aggregated data as data for one hour, processes the
attribute-aggregated data as time-aggregated data, and outputs the
time-aggregated data to the characteristic point extraction unit 48 (520).
[0078]
Next, the characteristic point extraction unit 48 extracts "characteristic
points" as its characteristic values from the input time-aggregated data and outputs
the time-aggregated data together with the characteristic points to the data
compression unit 50 (521 ).
[0079]
Next, the data compression unit 50 compresses the input time-aggregated
data and outputs the compressed data and the characteristic points to the
aggregated data write unit 52 (522).
[0080]
Next, the aggregated data write unit 52 receives the time-aggregated data
and data of the characteristic point, writes the received time-aggregated data and
characteristic point data to the aggregated data table 32 of the external storage
device 18 via the external storage interface 24 (523), and terminates the processing
in this routine.
[0081]
Under this circumstance, the aggregated data table 32 stores the
hourly-based time-aggregated data together with the data of the characteristic
points and the data aggregation time zone by associating them with the aggregation
10 (the number of vibrations). Furthermore, since the compressed data are
accumulated in the aggregated data table 32, the aggregated data can be
19
accumulated in a smaller data amount than the amount of data which are not
compressed.
[0082]
Next, the time series data analysis processing will be explained in
accordance with a flowchart of Fig. 10.
[0083]
This processing is started by activation of the data analysis unit 36 and the
data acquisition unit 38 of the time series data processing program 30 by the
processor 26.
[0084]
Firstly, the analysis reception unit 60 receives the analysis query 170 (S31);
the data acquisition unit 38 executes processing for creating lists of the analysis
target ID, the search target ID, the acquisition target ID, and the acquisition target
aggregation ID (S32); and then, the data acquisition unit 38 executes data
acquisition processing (S33). Subsequently, the analysis execution unit 62 executes
data extraction and analysis processing according to a condition and an attribute
(S34), executes processing for sending data accumulated in an analysis result
buffer (not shown) as result data to the client computer 12 (S35), and terminates the
processing in this routine.
[0085]
Next, the processing for creating the lists of the analysis target ID, the
search target ID, the acquisition target ID, and the acquisition target aggregation ID
will be explained in accordance with a flowchart in Fig. 11.
[0086]
This processing is the processing executed in step S32 in Fig. 1 0; and the
analysis reception unit 60 firstly creates an analysis target ID list from the selected
items (select_items) 174 of the analysis query 70 and writes the predicted number of
vibrations and the actual number of vibrations as the analysis target ID in this list
(S41).
[0087]
20
Next, the analysis reception unit 60 creates a search target ID list from the
search condition (where_condition) 178 of the analysis query 170 and writes the
predicted number of vibrations as the search target ID in this list (S42).
[0088]
Then, the analysis reception unit 60 creates an acquisition target ID list by
combining the analysis target ID and the search target ID and writes the predicted
number of vibrations and the actual number of vibrations as the acquisition target ID
in this list (S43).
[0089]
Next, the analysis reception unit 60 starts loop processing of the acquisition
target ID list from step S44 to step S48.
[0090]
Firstly, the data analysis reception unit 60: compares the acquisition target
I D with the attribute 122 of the attribute-aggregated information 90 and obtains the
"number of vibrations" as the aggr~gation ID which becomes an acquisition target
(S45); and judges whether or not the "number of vibrations" which is the aggregation
I D and becomes the acquisition target exists in the acquisition target aggregation I D
list (S46). If the "number of vibrations" does not exist, the data analysis reception
unit 60 adds the aggregation ID which becomes the acquisition target to the
acquisition target aggregation ID list (S47); and if the aggregation ID which becomes
the acquisition target already exists in the acquisition target aggregation ID list, the
data analysis reception unit 60 terminates the processing in this routine.
[0091]
Next, the data acquisition processing will be explained in accordance with a
flowchart in Fig. 12.
[0092]
This processing is the processing executed in step S33 in Fig. 10. Firstly,
the read time zone narrowing-down unit 70 receives the acquisition target
aggregation ID list from the analysis reception unit 60 (S51) and obtains, for
example, "07:20:00 to 08:40:00" as a data acquisition target time zone from a time
21
zone of the data acquisition target time zone (from_timerange) 176 of the analysis
query 170 based on the analysis query 170 (S52).
[0093]
Next, the read time zone narrowing-down unit 270 narrows down the data
acquisition time zone to a time zone where data satisfying the search condition
(where_condition) 178 of the analysis query 170 exists, by using the characteristic
points of the aggregated data table 32 (S53). For example, when the read time zone
narrowing-down unit 270 refers to the aggregated data table 32 in Fig. 7 and if time
series data whose predicted number of vibrations is 40 or more does not exist in a
first entry, but exists in only a second entry, the read time zone narrowing-down unit
270 narrows down the data acquisition target time zone from 7:20 to 8:40 to 8:00 to
8:40.
[0094]
Subsequently, the read time zone narrowing-down unit 70 outputs
information about the narrowed-down data acquisition target time zone to the
aggregated data read unit 72.
[0095]
Subsequently, from step S54 to S58, loop processing of the acquisition
target aggregation ID list is executed.
[0096]
Firstly, the aggregated data read unit 72 refers to the aggregated data table
32, obtains aggregated data belonging to the narrowed-down data acquisition target
time zone as aggregated data relating to the acquisition target aggregation ID from
the aggregated data table 32, and outputs the obtained aggregated data to the data
unzipping unit 7 4 (S55).
[0097]
Next, the data unzipping unit 74 unzips the input aggregated data and
outputs the unzipped aggregated data to the data time extraction unit 76 (S56).
[0098]
Then, the data time extraction unit 76 refers to the number of
22
time-aggregated data pieces 42 of the time-aggregated information 92 and extracts
an attribute-aggregated data list belonging to the narrowed-down data acquisition
target time zone based on the number of time-aggregated data pieces (S57); and on
condition that the entire processing regarding the acquisition target aggregation ID
list is completed, the data time extraction unit 76 terminates the processing in this
routine.
[0099]
Next, the data extraction and analysis processing according to the condition
and the attribute will be explained in accordance with a flowchart in Fig. 13.
[0100]
This processing is the processing in step S34 in Fig. 10. Firstly, the data
narrowing-down unit 78 obtains the data narrowing-down condition (the predicted
number of vibrations is 40 or more) from the condition 178 of the analysis query 170
(S61) and obtains the attribute-aggregated data list extracted by the data time
extraction unit 76 (S62).
[0101]
Subsequently, from step S63 to S68, loop processing of the
attribute-aggregated data list is executed.
[0102]
Firstly, the data narrowing-down unit 78 judges whether the
attribute-aggregated data satisfies the data narrowing-down condition or not (S64).
Specifically speaking, the data narrowing-down unit 78 judges whether or not any
data whose predicted number of vibrations is 40 or more exist in the
attribute-aggregated data. In other words, the data narrowing-down unit 78 judges
whether or not the value of the predicted number of vibrations is equal to the
reference value = 40 or more which becomes a condition of the search .target. If any
data whose predicted number of vibrations is 40 or more exist in the
attribute-aggregated data, the data narrowing-down unit 78 extracts combinations of
the time series data, regarding which the value of the predicted number of vibrations
is equal to the reference value = 40 or more which becomes the condition of the
23
search target, as combinations of the narrowed-down time series data.
[01 03]
If any data whose predicted number of vibrations is 40 or more exist in the
attribute-aggregated data, the data attribute extraction unit 78 extracts data of the
analysis target ID list from the combinations of the time series data
(attribute-aggregated data) narrowed down by the data narrowing-down unit 78
(S65). Specifically speaking, the data attribute extraction unit 78 extracts sets of the
time series data relating to the predicted number of vibrations and the actual number
of vibrations, which are specified by the selected items 174 of the analysis query
170, one set at a time as a plurality of pieces of time series data which are the
analysis targets, from the attribute-aggregated data.
[0104]
The data attribute extraction unit 80 outputs each combination of the plural
pieces of the extracted time series data as the plurality of pieces of time series data,
which are the analysis targets, to the analysis execution unit 62.
[0105]
The analysis execution unit 62 analyzes the plural sets of the time series
data extracted by the data attribute extraction unit 80 as specified by the selected
items 174 of the analysis query 170 (S66). Specifically speaking, the analysis
execution unit 62 analyzes on the sets of the plural pieces of time series data which
become the analysis targets and whose predicted number of vibrations is 40 or
more to find the divergence frequency by subtracting the actual number of vibrations
from the predicted number of vibrations.
[01 06]
Subsequently, the analysis execution unit 62 accumulates each set of the
analysis results in the analysis result buffer (not shown) (S67) and terminates the
processing in this routine.
[01 07]
According to this embodiment, a plurality of pieces of time series data
generated in the same cycle are gathered and aggregated based on the attribute
24
(the number of vibrations) and also gathered and aggregated on the set time basis,
so that the plurality of pieces of time series data generated in the same cycle can be
accumulated efficiently.
[0108]
Furthermore, according to this embodiment, combinations of the time series
data generated in the same cycle are extracted as the time series data to be used
for analysis from the aggregated data table 32 by accessing the aggregated data
table 32, in which the plurality of pieces of time series data are accumulated based
on the attribute (the number of vibrations), based on the attribute, so that the time
series data used for the analysis can be efficiently accessed and analyzed.
Embodiment 2
[0109]
This embodiment is designed so that: upon data accumulation,
combinations of time series data generated in different cycles (time series data
relating to power generation and power consumption) are selected as combinations
of a plurality of pieces of time series data, which become analysis targets, a plurality
of sets of the selected combinations of the time series data are gathered and
aggregated on a set time basis, and the plurality of aggregated sets of the time
series data are accumulated in a storage unit by associating them with an attribute
(electric power); and upon data analysis, the combinations of time series data
generated in the different cycles (time series data relating to the power generation
and the power consumption) are extracted, as time series data to be used for the
analysis, from the storage unit.
[0110]
Fig. 14 shows the overall configuration of a second embodiment according
to the present invention.
[0111]
With the time series data processing unit 16 for a computer system
according to this embodiment, a data cycle aggregation unit 43 is located between
the data reception unit 42 and the data attribute aggregation unit 44 for the data
25
accumulation unit 34, an attribute buffer 53 is located instead of the attribute buffer
54, a time buffer 55 is located instead of the time buffer 56, a cycle buffer 57 is
located as a buffer managed by the data cycle aggregation unit 43, a data cycle
extraction unit 82 is located after the data attribute extraction unit 80 for the data
acquisition unit 38; in the setting information storage area 40, an
attribute-aggregated information 91 is located instead of the attribute-aggregated
information 90, time-aggregated information 93 is located instead of the
time-aggregated information 92, and cycle-aggregated information 95 is newly
located; and furthermore, in the external storage device 18, an aggregated data
table 33 is stored instead of the aggregated data table 32; and the time series data
processing unit 16 is designed to process time series data generated in different
cycles as the time series data; and other components are the same as those in the
first embodiment.
[0112]
The data cycle aggregation unit 43 processes time series data whose
generation cycles are different, among the time series data output from the data
reception unit 42, by gathering them as time series data for one hour, so that for
example, time series data generated in a one-hour cycle (time series data relating to
the power generation) and time series data generated in a one-minute cycle (time
series data relating to the power consumption) are accumulated as accumulated
data in the cycle buffer 57 and each of the time series data generated in the
one-hour cycle and the time series data generated in the one-minute cycle is output
to the data attribute aggregation unit 44.
[0113]
The data attribute aggregation unit 44: combines a plurality of pieces of time
series data, which become analysis targets, that is, a plurality of pieces of time
series data generated in different cycles, among the time series data which have
been input; combines the combined plural pieces of time series data, for example,
the time series data relating to the power generation and the timer series data
relating to the power consumption; aggregates the combined plural pieces of time
26
series data by associating them with a mutually related attribute, for example,
electric power (hereinafter sometimes referred to as attribute aggregation); and
accumulates the aggregated time series data as attribute-aggregated data in the
attribute buffer 53 and outputs them to the data time aggregation unit 46.
[0114]
The data time aggregation unit 46 collects the attribute-aggregated data,
which have been input, only for a set time period and processes them as
time-aggregated data. For example, the data time aggregation unit 46 aggregates
the attribute-aggregated data, which have been input, on a 24-hour basis,
accumulates them as time-aggregated data in the time buffer 55, and outputs them
to the characteristic point extraction unit 48.
[0115]
The characteristic point extraction unit 48 extracts characteristic points from
the time-aggregated data, which have been input, and outputs the extracted
characteristic points as well as the input time-aggregated data to the data
compression unit 50.
[0116]
The data compression unit 50 compresses the time-aggregated data, which
have been output from the characteristic point extraction unit 48, and outputs the
compressed data and the characteristic points, which have been output from the
characteristic point extraction unit 48, to the aggregated data write unit 52.
[0117]
The aggregated data write unit 52 writes the data, which have been output
from the data compression unit 50, as aggregated data to the aggregated data table
32 of the external storage device 18.
[0118]
In the same manner as in the first embodiment, the analysis reception unit
60 for the data analysis unit 36 receives an analysis query from the client computer
12 and outputs the received analysis query to the data acquisition unit 38. The
analysis execution unit 62 executes analysis as specified by the analysis query
27
based on the processing results of the data acquisition unit 38.
[0119]
The read time zone narrowing-down unit 70, the aggregated data read unit
72, the data unzipping unit 74, the data time extraction unit 76, the data
narrowing-down unit 78, and the data attribute extraction unit 80 for the data
acquisition unit 38 execute processing relating to the time series data in the same
manner as in the first embodiment.
[0120]
Under this circumstance, the data attribute extraction unit 80: gathers and
extracts combinations of time series data relating to one piece of power generation
and time series data relating to 60 pieces of the power consumption for each
analysis target as the plurality of pieces of time series data which become the
analysis targets and whose aggregation ID is associated with the electric power,
among the attribute-aggregated data narrowed down by the data narrowing-down
unit 78; and outputs each of the plural pieces of the extracted time series data to the
analysis execution unit 62.
[0121]
Next, Fig. 15 shows the structure of the cycle-aggregated information 95.
[0122]
The cycle-aggregated information 95 is constituted from an attribute 200
and the number of cycle-aggregated data pieces 202. The attribute 200 stores, for
example, "power generation" and "power consumption" as the attribute of the time
series data which are the analysis targets. The number of cycle-aggregated data
pieces 202 stores, for example, information indicating "one piece" corresponding to
the power generation and "60 pieces" corresponding to the power consumption as
the number of necessary data pieces when performing analysis on an hourly basis.
This is because when analyzing the difference of electric power per hour, one piece
of data is used as the power generation and 60 pieces of data are used as the
power consumption for one minute.
[0123]
28
Fig. 16 shows the structure of the cycle buffer 57.
[0124]
An information storage area of the cycle buffer 57 is configured in a table
format and the table is constituted from a date and time field 210, an attribute field
212, and an accumulated data field 214.
[0125]
Each entry of the date and time field 210 stores information about a data
and time when accumulating data in the cycle buffer 57, as a numerical value
together with the relevant date. Each entry of the attribute field 212 stores, for
example, information indicating the "power generation" or the "power consumption"
as the attribute of the time series data accumulated in the cycle buffer 57. Each
entry of the accumulated data field 214 stores the value of the time series data
(accumulated data) accumulated in the cycle buffer 57.
[0126]
For example, "240" is stored as the value of the time series data 04 in Fig. 2
corresponding to the power generation and "6.7," "7.1 ,"and "12.4" are stored as the
values of the time series data 03, 07, 08 in Fig. 2 corresponding to the power
consumption.
[0127]
Next, Fig. 17 shows the structure of the attribute-aggregated information 91.
[0128]
The attribute-aggregated information 91 is constituted from an aggregation
10 220 and an attribute 222. The attribute 222 stores, for example, "power
generation" and "power consumption" as an attribute of time series data generated
in different cycles. The aggregation 10 220 stores "electric power'' as an attribute
indicated by the power generation and the power consumption stored in the attribute
222 and as a mutually related attribute.
[0129]
Next, Fig. 18 shows the structure of the attribute buffer 53.
[0130]
29
An information storage area of the attribute buffer 53 is configured in a table
format and the table is constituted from a cycle-aggregated time zone field 230, an
attribute field 232, and a cycle-aggregated data field 234.
[0131]
Each entry of the cycle-aggregated time zone field 230 stores information
about a cycle-aggregated time zone on a hourly basis. Each entry of the attribute
field 232 stores, for example, "power generation" and "power consumption" as the
attribute of a plurality of pieces of time series data which become analysis targets.
[0132]
Each entry of the cycle-aggregated data field 234 stores, for example, one
value of the power generation corresponding to the power generation and 60 values
of the power consumption collected in a one-minute cycle corresponding to the
power consumption.
[0133]
Next, Fig. 19 shows the structure of the time-aggregated information 93.
[0134]
The time-aggregated information 93 is constituted from an aggregation 10
240 and the number of time-aggregated data pieces 242. The aggregation 10 240
stores, for example, "electric power'' as an attribute mutually related to the "power
generation" and the "power consumption". The number of time-aggregated data
pieces 242 stores, for example, "24 pieces" in order to aggregate time of electric
power data on a 24-hour basis.
[0135]
Next, Fig. 20 shows the structure of the time buffer 55.
[0136]
The time buffer 55 is constituted from a data aggregation time zone 250 and
attribute-aggregated data 252. The data aggregation time zone 250 stores
information about a time zone to aggregate data as 24-hour based information. The
attribute-aggregated data 252 stores data gathered for each hour. For example, in a
case of the time series data in Fig. 2, the time series data 04, 03, 07, and so on up
30
to 08 are stored as data for one hour, which become analysis targets.
[0137]
Next, Fig. 21 shows the structure of the aggregated data table 33.
[0138]
An information storage area of the aggregated data table 33 is configured in
a table format and the table is constituted from an aggregation 10 field 260, a
characteristic point field 262, a data aggregation time zone field 264, and a
time-aggregated data field 266.
[0139]
Each entry of the aggregation 10 field 260 stores, for example, "electric
power'' as an attribute aggregating the power generation and the power
consumption.
[0140]
Each entry of the characteristic point field 262 stores information about
characteristic points of time series collected every 24 hours. For example, a
"maximum value of power generation: 300" is stored as a maximum value of the
power generation and a "minimum value of power generation: 200" is stored as a
minimum value of the power generation. Furthermore, a "maximum value of power
consumption: 1 0" is stored as a maximum value of the power consumption and a
"minimum value of power consumption: 5" is stored as a minimum value of the
power consumption.
[0141]
Each entry of the data aggregation time zone field 264 stores, for example,
information about a data aggregation time zone to obtain the power generation and
the power consumption for 24 hours.
[0142]
The time-aggregated data field 266 stores a value related to
cycle-aggregated data on an hourly basis as data for 24 hours. For example, in a
case of the time series data in Fig. 2, respective values of the time series data 04,
03, 07, and so on up to 08 are stored as data for one hour.
31
[0143]
Next, Fig. 22 shows the structure of the analysis query 270.
[0144]
The analysis query 270 is constituted from a selected range (select_range)
272, selected items (select_items) 274, a data acquisition target time zone
(from_timerange) 276, and a search condition (where_condition) 278.
[0145]
The selected range (select_range) 272 stores one hour as a processing
time unit for the data analysis unit 36 to analyze data.
[0146]
The selected items (select_items) 274 store, as an analysis target ID, for
example, Power Generation - SUM (Power Consumption) AS Electric Power
Difference.
[0147]
The data acquisition target time zone (from_timerange) 276 stores, for
example "201 0-05-01 07:00:00" to "201 0-05-01 17:00:00" as information about the
data acquisition target time zone.
[0148]
The condition (where_condition) 278 stores power generation >= 250 as a
condition to become the search target I D.
[0149]
Next, the time series data accumulation processing will be explained in
accordance with a flowchart in Fig. 23.
[0150]
This processing is started by activation of the data accumulation unit 34 of
the time series data processing program 30 by the processor 26.
[0151]
Firstly, the data reception unit 42 receives the time series data, which have
been output from the time series data source 10, via the network 14 and the
communications interface 22 and delivers the received time series data to the data
32
cycle aggregation unit 43 (S71).
[0152]
Next, the data cycle aggregation unit 43 sequentially accumulates the input
time series data in the cycle buffer 57 (S72) and judges whether or not the number
of accumulated data pieces in the cycle buffer 57 exceeds the number specified by
the number of cycle-aggregated data pieces 202 of the cycle-aggregated
information 95 (S73); and if the number of data pieces accumulated in the cycle
buffer 57 does not exceed the number of pieces specified by the number of
cycle-aggregated data pieces 202, the data cycle aggregation unit 43 returns to the
processing in step S71; and if the number of data pieces accumulated in the cycle
buffer 57 exceeds the number of pieces specified by the number of
cycle-aggregated data pieces 202, the data cycle aggregation unit 43 proceeds to
processing in step S7 4.
[0153]
In step S73, the data cycle aggregation unit 43: judges whether or not one
piece of time series data relating to the power generation is accumulated during the
process of sequentially accumulating the input time series data in the cycle buffer
57; and also judges whether or not the number of time series data pieces relating to
the power consumption has reached 60 pieces.
[0154]
Next, the data cycle aggregation unit 43: obtains the accumulated data
accumulated in the cycle buffer 57 (data including one piece of the time series data
relating to the power generation and 60 pieces of the time series data relating to the
power consumption) from the cycle buffer 57 and then deletes the data from the
cycle buffer 57 (S74); executes cycle aggregation to gather the accumulated data
obtained from the cycle buffer 57 as data in a one-hour cycle and outputs the
cycle-aggregated data on which the cycle aggregation was executed (data including
the time series data relating to the power generation and the time series data
relating to the power consumption) to the data attribute aggregation unit 44 (S75).
[0155]
33
Next, the data attribute aggregation unit 44 receives the cycle-aggregated
data and sequentially accumulates these cycle-aggregated data in the attribute
buffer 53 corresponding to the aggregation ID (576) and judges whether or not the
cycle-aggregated data of all attributes (the power generation and the power
consumption) on the same date and time exist in the attribute buffer 53
corresponding to the aggregation ID (577); and if it is determined that the
cycle-aggregated data of all the attributes on the same date and time do not exist,
the data attribute aggregation unit 44 returns to the processing in step 571; and if it
is determined that the cycle-aggregated data of all the attributes on the same date
and time exist, the data attribute aggregation unit 44 obtains the cycle-aggregated
data of all the attributes (the power generation and the power consumption) on the
same date and time from the attribute buffer 53 corresponding to the aggregation I D
and then deletes the data from the attribute buffer 53 (578).
[0156]
Next, the data attribute aggregation unit 44 combines the cycle-aggregated
data of all the attributes (the power generation and the power consumption) on the
same date and time, which are obtained from the attribute buffer 53, for each of the
plural pieces of time series data which become the analysis targets, performs
attribute aggregation to associate the cycle-aggregated data with the aggregation ID
(electric power), processes the combined plural pieces of the time series data as
attribute-aggregated data, and outputs these attribute-aggregated data to the data
time aggregation unit 46 (579).
[0157]
Next, the data time aggregation unit 46 receives the attribute-aggregated
data and sequentially accumulates these attribute-aggregated data in the time buffer
55 (580}, and judges whether the number of the attribute-aggregated data pieces
accumulated in the time buffer 55 exceeds the number of pieces stored in the
number of time-aggregated data pieces 242 of the time-aggregated information 93,
for example, 24 pieces (581); and if the number of the attribute-aggregated data
pieces in the time buffer 55 does not exceed the number of time-aggregated data
34
pieces, the data time aggregation unit 46 returns to the processing in step 871 to
collect the attribute-aggregated data for 24 hours; and if the number of the
attribute-aggregated data pieces in the time buffer 55 exceeds the number of
time-aggregated data pieces, the data time aggregation unit 46 proceeds to
processing in step 882, recognizing that the attribute-aggregated data for 24 hours
have been collected.
[0158]
Next, in step 882, the data time aggregation unit 46 obtains all pieces of the
attribute-aggregated data from the time buffer 55 corresponding to the aggregation
ID (electric power) and then deletes the data (attribute-aggregated data) from the
time buffer 55.
[0159]
Next, the data time aggregation unit 46 executes time aggregation to gather
all the pieces of the obtained attribute-aggregated data as data for 24 hours,
processes the attribute-aggregated data, on which the time aggregation was
executed, as time-aggregated data, and outputs the time-aggregated data to the
characteristic point extraction unit 48 (883).
[0160]
Next, the characteristic point extraction unit 48 extracts "characteristic
points" as its characteristic value from the input time-aggregated data and outputs
the time-aggregated data together with the extracted characteristic points to the
data compression unit 50 (884).
[0161]
Next, the data compression unit 50 compresses the input time-aggregated
data and outputs the compressed data and the characteristic points to the
aggregated data write unit 52 (885).
[0162]
Next, the aggregated data write unit 52 receives the time-aggregated data
and data of the characteristic point, writes the received time-aggregated data and
characteristic point data to the aggregated data table 33 of the external storage
35
device 18 via the external storage interface 24 (886}, and terminates the processing
in this routine. Under this circumstance, the aggregated data table 33 stores the
24-hour based time-aggregated data together with the characteristic points and data
of the data aggregation time zone by associating them with the aggregation ID
(electric power).
[0163]
Next, the time series data analysis processing will be explained in
accordance with a flowchart in Fig. 24.
[0164]
This processing is executed by the data analysis unit 36 and the data
acquisition unit 38. Firstly, the analysis reception unit 60 receives the analysis query
270 issued from the client computer 12 (891 ); and then, the data acquisition unit 38
executes processing for creating lists of the acquisition target ID, the search
condition ID, the acquisition target ID, and the acquisition target aggregation ID
(892).
[0165]
Next, the data acquisition unit 38 executes data acquisition processing
(893); then, the analysis execution unit 62 executes data extraction and analysis
processing according to a condition, an attribute, and a cycle (894); and lastly, the
analysis execution unit 62 sends result data accumulated in the analysis result
buffer to the client computer 12 (895) and terminates the processing in this routine.
[0166]
Incidentally, the content of the processing for creating the lists of the
acquisition target ID, the search condition ID, the acquisition target ID, and the
acquisition target aggregation I D in step 892 is the same as that of the processing in
Fig. 11, except that it is processing based on the analysis query 270; and
furthermore, the content of the data acquisition processing in step 893 is the same
as that of the processing in Fig. 12, except that it is based on the analysis query 270.
Therefore, an explanation about them has been omitted.
[0167]
36
Next, the data extraction and analysis processing according to the condition,
attribute, and cycle will be explained in accordance with a flowchart in Fig. 25.
[0168]
This step is the processing executed in step S94 in Fig. 24. Firstly, if the
data narrowing-down condition is set so that, for example, the power generation is
equal to a reference value = 250 or more, the data narrowing-down unit 78 obtains
the power generation 250 or more from the search condition (where_condition) 278
of the analysis query 270 (S 101) and then obtains the attribute-aggregated data list
extracted by the data time extraction unit 76 (S 1 02).
[0169]
Subsequently, loop processing based on the attribute-aggregated data list is
executed from step S 103 to S 1 09.
[0170]
Firstly, the data narrowing-down unit 78 judges whether the obtained
attribute-aggregated data satisfy the data narrowing-down condition or not (S104).
Specifically speaking, the data narrowing-down unit 78 judges whether or not any
data whose power generation is 250 or more exist in the attribute-aggregated data;
and if any data whose power generation is 250 or more exist in the
attribute-aggregated data, the data narrowing-down unit 78 proceeds to step S105;
and if any data whose power generation is 250 or more do not exist in the
attribute-aggregated data, the data narrowing-down unit 78 proceeds to processing
in step S109.
[0171]
In step S105, the data attribute extraction unit 80 extracts the
attribute-aggregated data of the analysis target 10 list from the attribute-aggregated
data. Specifically speaking, the data attribute extraction unit 80 extracts the
attribute-aggregated data specified in the selected items 27 4 of the analysis query
270, for example, attribute-aggregated data including one piece of the time series
data relating to the power generation and 60 pieces of the time series data relating
to the power consumption as data for one hour.
37
[0172]
Next, the data cycle extraction unit 82 extracts data (data accumulated
during a period from 7:00:00 to 17:00:00) within the data acquisition target time zone
(from_timerange) 276 of the analysis query 270 based on the number of data pieces
specified by the number of cycle-aggregated data pieces 202 of the
cycle-aggregpted information 95 (power generation: 1 piece; and power
consumption: 60 pieces) (S106).
[0173]
Subsequently, the analysis execution unit 62 analyzes the data extracted by
the data cycle extraction unit 82 with respect to the selected items (select_items)
274 of the analysis query 270. For example, if the selected items (select_items) 274
are "Power Generation- SUM (Power Consumption) AS Electric Power Difference,"
the analysis execution unit 62 executes an operation to calculate the electric power
difference by subtracting the power consumption for one hour from the power
generation.
[0174]
Next, the analysis execution unit 62 accumulates the analysis results in an
analysis result buffer (not shown) (S108); and on condition that the execution of the
entire processing on the attribute-aggregated data list is completed, the analysis
execution unit 62 terminates the processing in this routine.
[0175]
According to this embodiment, a plurality of pieces of time series data
generated in different cycles are gathered and aggregated based on the attribute
(electric power) and also gathered and aggregated on a set time basis, so that the
plurality of pieces of time series data generated in the different cycles can be
accumulated efficiently.
[0176]
Furthermore, according to this embodiment, combinations of the time series
data generated in the different cycles are extracted as the time series data to be
used for analysis from the aggregated data table 33, in which the plurality of pieces
38
of time series data are accumulated based on the attribute (electric power), by
accessing the aggregated data table 33 based on the attribute. So, the plurality of
pieces of time series data used for the analysis can be efficiently accessed and
analyzed.
Reference Signs List
[0177]
10 time series data source, 12 client computer, 14 network, 16 time series
data processing unit, 18 external storage device, 20 memory, 26 processor, 30 time
series data processing program, 32 aggregated data table, 34 data accumulation
unit, 36 data analysis unit, 38 data acquisition unit, 40 setting information storage
area , 42 data reception unit, 43 data cycle aggregation unit , 44 data attribute
aggregation unit, 46 data time aggregation unit, 48 characteristic point extraction
unit, 50 data compression unit, 52 aggregated data write unit, 60 analysis reception
unit, 62 analysis execution unit, 70 read time zone narrowing-down unit, 72
aggregated data read unit, 7 4 data unzipping unit, 76 data time extraction unit, 78
data narrowing-down unit, 80 data attribute extraction unit, and 82 data cycle
extraction unit.

WE CLAIM:
[Claim 1]
A time series data processing unit comprising a data processing unit for
sequentially inputting and processing time series data from a time series data
generation source, a storage unit for accumulating a processing result of the data
processing unit, and a data acquisition analysis unit for obtaining data from the
storage unit and analyzing the obtained data according to an analysis request,
wherein the data processing unit combines a plurality of pieces of time
series data, which become analysis targets, among the time series data, which have
been input, and accumulates the combined plural pieces of time series data In the
storage unit by associating them with a mutually related attribute; and
wherein if a search target specified by the analysis request is the attribute,
the data acquisition analysis unit accesses the storage unit by using the attribute as
the search target, extracts the plurality of pieces of time series data, which are time
series data corresponding to the attribute and become the analysis targets, from the
storage unit, and executes analysis as specified by the analysis request by using the
extracted plural pieces of time series data.
[Claim 2]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in a same cycle as combinations of the plurality of pieces of time
series data, which become the analysis targets, gathers and aggregates plural sets
of the selected combinations of time series data on a set time basis, and
accumulates the aggregated plural sets of time series data in the storage unit by
associating them with the attribute; and
wherein the data acquisition analysis unit extracts the combinations of time
series data generated in the same cycle as time series data to be used for the
analysis from the storage unit.
[Claim 3]
The time series data processing unit according to claim 1,
40
wherein the data processing unit selects combinations of the time series
data generated in the same cycle as combinations of the plurality of pieces of time
series data, which become the analysis targets, and accumulates the selected
combinations of time series data in the storage unit by associating them with the
attribute; and
wherein the data acquisition analysis unit extracts the combinations of time
series data generated in the same cycle as time series data to be used for the
analysis from the storage unit.
[Claim 4]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in different cycles as combinations of the plurality of pieces of time
series data, which become the analysis targets, and accumulates the selected
combinations of time series data in the storage unit by associating them with the
attribute; and
wherein the data acquisition analysis unit extracts the combinations of time
series data generated in the different cycles as time series data to be used for the
analysis from the storage unit.
[Claim 5]
The time series data processing unit according to claim 3,
wherein the data processing unit gathers and aggregates plural sets of the
selected combinations of time series data on a set time basis and accumulates the
aggregated plural sets of time series data in the storage unit by associating them
with the attribute; and
wherein the data acquisition analysis unit extracts the aggregated plural
sets of time series data as time series data to be used for the analysis from the
storage unit.
[Claim 6]
The time series data processing unit according to claim 4,
wherein the data processing unit gathers and aggregates plural sets of the
41
selected combinations of time series data on a set time basis and accumulates the
aggregated plural sets of time series data in the storage unit by associating them
with the attribute; and
wherein the data acquisition analysis unit extracts the aggregated plural
sets of time series data as time series data to be used for the analysis from the
storage unit.
[Claim 7]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in the same cycle as combinations of the plurality of pieces of time
series data, which become the analysis targets, extracts a characteristic point of
each piece of the time series data belonging to the selected combinations of time
series data, and accumulates the selected combinations of time series data together
with the extracted characteristic point in the storage unit by associating them with
the attribute; and
wherein if a reference value for the characteristic point of the time series
data belonging to the combinations of time series data generated in the same cycle
is specified by the analysis query as a condition of the search target, the data
acquisition analysis unit extracts the time series data, whose characteristic point
satisfies the reference value, as the time series data to be used for the analysis
among the combinations of time series data generated in the same cycle from the
storage unit.
[Claim 8]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in different cycles as combinations of the plurality of pieces of time
series data, which become the analysis targets, extracts a characteristic point of
each piece of the time series data belonging to the selected combinations of time
series data, and accumulates the selected combinations of time series data together
with the extracted characteristic point in the storage unit by associating them with
42
the attribute; and
wherein if a reference value for the characteristic point of the time series
data belonging to the combinations of time series data generated in the different
cycles is specified by the analysis query as a condition of the search target, the data
acquisition analysis unit extracts the time series data, whose characteristic point
satisfies the reference value, as the time series data to be used for the analysis
among the combinations of time series data generated in the different cycles from
the storage unit.
[Claim 9]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in the same cycle as combinations of the plurality of pieces of time
series data, which become the analysis targets, compresses each piece of time
series data belonging to the selected combinations of time series data, and
accumulates the compressed combinations of time series data in the storage unit by
associating them with the attribute; and
wherein the data acquisition analysis unit extracts the compressed
combinations of time series data as time series data to be used for the analysis,
among the combinations of time series data generated in the same cycle, from the
storage unit and unzips each extracted piece of the time series data.
[Claim 10]
The time series data processing unit according to claim 1,
wherein the data processing unit selects combinations of the time series
data generated in different cycles as combinations of the plurality of pieces of time
series data, which become the analysis targets, compresses each piece of time
series data belonging to the selected combinations of time series data, and
accumulates the compressed combinations of time series data in the storage unit by
associating them with the attribute; and
wherein the data acquisition analysis unit extracts the compressed
combinations of time series data as time series data to be used for the analysis,
43
among the combinations of time series data generated in the different cycles, from
the storage unit and unzips each extracted piece of the time series data.
[Claim 11]
A time series data processing method comprising a data processing unit for
sequentially inputting and processing time series data from a time series data
generation source, a storage unit for accumulating a processing result of the data
processing unit, and a data acquisition analysis unit for obtaining data from the
storage unit and analyzing the obtained data according to an analysis request,
the time series data processing method comprising:
a step executed by the data processing unit combining a plurality of pieces
of time series data, which become analysis targets, among the time series data,
which have been input, and accumulating the combined plural pieces of time series
data in the storage unit by associating them with a mutually related attribute;
a step executed, if a search target specified by the analysis request is the
attribute, by the data acquisition analysis unit accessing the storage unit by using
the attribute as the search target;
a step executed by the data acquisition analysis unit extracting the plurality
of pieces of time series data, which are time series data corresponding to the
attribute and become the analysis targets, from the storage unit; and
a step executed by the data acquisition analysis unit executing analysis
specified by the analysis request by using the plural pieces of time series data
extracted in the above step.
[Claim 12]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in a same cycle as combinations of the plurality of pieces
of time series data, which become the analysis targets;
a step executed by the data processing unit gathering and aggregating
plural sets of the selected combinations of time series data on a set time basis;
a step executed by the data processing unit accumulating the plural sets of
44
time series data aggregated in tlie above step in the storage unit by associating
them with the attribute;
a step executed by the data acquisition analysis unit extracting the
combinations of time series data generated in the same cycle as time series data to
be used for the analysis from the storage unit.
[Claim 13]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in the same cycle as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit accumulating the combinations
of time series data selected in the above step in the storage unit by associating them
with the attribute; and
a step executed by the data acquisition analysis unit extracting the
combinations of time series data generated in the same cycle as time series data to
be used for the analysis from the storage unit.
[Claim 14]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in different cycles as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit accumulating the combinations
of time series data selected in the above step in the storage unit by associating them
with the attribute; and
a step executed by the data acquisition analysis unit extracting the
combinations of time series data generated in the different cycles as time series
data to be used for the analysis from the storage unit.
[Claim 15]
The time series data processing method according to claim 13, comprising:
a step executed by the data processing unit gathering and aggregating
45
plural sets of the selected combinations of time series data on a set time basis;
a step executed by the data processing unit accumulating the plural sets of
time series data aggregated in the above step in the storage unit by associating
them with the attribute; and
a step executed by the data acquisition analysis unit extracting the plural
sets of time series data aggregated in the above step as time series data to be used
for the analysis from the storage unit.
[Claim 16]
The time series data processing method according to claim 14, comprising:
a step executed by the data processing unit gathering and aggregating
plural sets of the selected combinations of time series data on a set time basis;
a step executed by the data processing unit accumulating the plural sets of
time series data aggregated in the above step in the storage unit by associating
them with the attribute; and
a step executed by the data acquisition analysis unit extracting the plural
sets of time series data aggregated in the above step as time series data to be used
for the analysis from the storage unit.
[Claim 17]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in the same cycle as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit extracting a characteristic point
of each piece of the time series data belonging to the combinations of time series
data selected in the above step;
a step executed by the data processing unit accumulating the combinations
of time series data selected in the above step together with the extracted
characteristic point in the storage unit by associating them with the attribute; and
a step executed, if a reference value for the characteristic point of the time
series data belonging to the combinations of time series data generated in the same
46
cycle is specified by the analysis query as a condition of the search target, by the
data acquisition analysis unit extracting the time series data, whose characteristic
point satisfies the reference value, as the time series data to be used for the
analysis among the combinations of time series data generated in the same cycle
from the storage unit.
[Claim 18]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in different cycles as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit extracting a characteristic point
of each piece of the time series data belonging to the combinations of time series
data selected in the above step;
a step executed by the data processing unit accumulating the combinations
of time series data selected in the above step together with the extracted
characteristic point in the storage unit by associating them with the attribute; and
a step executed, if a reference value for the characteristic point of the time
series data belonging to the combinations of time series data generated in the
different cycles is specified by the analysis query as a condition of the search target,
by the data acquisition analysis unit extracting the time series data, whose
characteristic point satisfies the reference value, as the time series data to be used
for the analysis among the combinations of time series data generated in the
different cycles from the storage unit.
[Claim 19]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in the same cycle as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit compressing each piece of time
series data belonging to the combinations of time series data selected in the above
47
step; and
a step executed by the data processing unit accumulating the combinations
of time series data compressed in the above step in the storage unit by associating
them with the attribute;
a step executed by the data acquisition analysis unit extracting the
compressed combinations of time series data as time series data to be used for the
analysis, among the combinations of time series data generated in the same cycle,
from the storage unit; and
a step executed by the data acquisition analysis unit unzipping each piece
of the time series data extracted in the above step.
[Claim 20]
The time series data processing method according to claim 11, comprising:
a step executed by the data processing unit selecting combinations of the
time series data generated in different cycles as combinations of the plurality of
pieces of time series data, which become the analysis targets;
a step executed by the data processing unit compressing each piece of time
series data belonging to the combinations of time series data selected in the above
step; and
a step executed by the data processing unit accumulating the combinations
of time series data compressed in the above step in the storage unit by associating
them with the attribute;
a step executed by the data acquisition analysis unit extracting the
compressed combinations of time series data as time series data to be used for the
analysis, among the combinations of time series data generated in the different
cycles, from the storage unit; and
a step executed by the data acquisition analysis unit unzipping each piece
of the time series data extracted in the above step.