JP6091873B2 - Domestic group behavior estimation device and household group behavior estimation method - Google Patents

Description

  Embodiments described herein relate generally to an intra-household group behavior estimation apparatus and an intra-household group behavior estimation method that estimate group behavior performed in a household at the same time zone.

  With the popularization of HEMS (Home Energy Management System), it is considered that power consumption control technology in an efficient and low-load form of energy consumption in each household is important. In such a power consumption control technique, a technique for estimating the amount of power in advance is important, and is proposed in Patent Documents 1 to 4, for example.

JP 2011-138492 A JP 2011-81431 A JP 2011-81540 A JP 2010-170419 A

  However, in order to grasp the behavior pattern of one household unit, a method for collecting data based on some characteristic of the behavior in the household is necessary. It is also important to calculate a power consumption trend based on the characteristics. Furthermore, since data collection in a form that describes behavior patterns and data related to the use of equipment for many households requires man-hours, a technique for analyzing the characteristics of household behavior while achieving the reduction in man-hours is required. is there.

  An object of the present embodiment is to provide an intra-household group behavior estimation apparatus and an intra-household group behavior estimation method capable of accurately extracting behavior patterns simultaneously executed by a plurality of members in a household.

  The intra-household group behavior estimation device according to the present embodiment matches the behavior label series representing the behavior pattern of each member constituting the household with the history data of the device used in the implementation of the behavior pattern, An action label using device conversion unit for converting to a label device use sequence, a feature using device sequence extracting unit for extracting a characteristic device use sequence from the behavior label using device sequence as a template for each member attribute, and sensing at each unit time A mixed use device state series separation unit that separates a mixed use device state sequence composed of a set of use device state sequences of the entire household as a use device state sequence of each member using the template, and a use device state of each member Device action label used to convert a series into a group action label set including an action label series with family attributes ; And a section.

The block diagram which shows the structure of the household group action estimation apparatus which concerns on this embodiment. The figure which shows an example of the data structure accumulate | stored in action label series teacher DB. The figure which shows an example of the data structure accumulate | stored in action label series teacher DB. The figure which shows an example of the data structure accumulate | stored in action label utilization apparatus DB. The flowchart which shows the process of an action label using apparatus conversion part. The figure which shows the example of extraction of the template classified by family attribute. The figure which shows the example of extraction of the template classified by family attribute. The figure which shows the example of extraction of the template classified by family attribute. The figure which shows an example of the data structure accumulate | stored in family mixed use apparatus state series DB. The figure which shows operation | movement of a mixing use apparatus state series separation part. The figure which shows operation | movement of a mixing use apparatus state series separation part. The figure which shows operation | movement of a mixing use apparatus state series separation part. The figure which shows an example of the selected template and group use apparatus status set. The flowchart which shows a use equipment label applicable location selection and evaluation function application order determination routine. The flowchart which shows the subroutine 1. The flowchart which shows an evaluation function application routine. The flowchart which shows the process of a group use apparatus state label collection extraction part. The figure which shows the specific example of operation | movement of a group use apparatus state label collection extraction part. The figure which shows use apparatus action label conversion processing. The figure which shows group simultaneous action label set extraction processing. The flowchart which shows a group simultaneous action label set extraction process. The figure which shows operation | movement of the power consumption calculation part at the time of a group action.

Hereinafter, the household group behavior estimation apparatus according to the present embodiment will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing the configuration of the household group behavior estimation apparatus according to the present embodiment. The household group behavior estimation apparatus includes a behavior label using device conversion unit 101, a feature using device series extraction unit 102, a mixed use device state series separation unit 103, a group use device state label set extraction unit 104, and a used device action label conversion unit 105. , And the group simultaneous action label set extraction unit 106 as a processing unit, the action label series teacher DB 108, the action label use device DB 109, the family attribute DB 111, the family mixed use device state series DB 113, and the action label DB 116 with a single action flag are stored in advance. By giving it, we estimate the group behavior to be executed in the same time zone in the household. Further, the household group behavior estimation apparatus includes a group behavior power consumption calculation unit 107, and estimates the amount of power consumed by the group behavior. The household group behavior estimation device can be realized by, for example, using a general-purpose computer device as basic hardware and configuring each processing unit as a control program.

  The action label series teacher DB 108 includes data describing household behavior patterns collected through questionnaires or the like as action label series. The behavior label series teacher DB 108 is a database that stores daily behavior patterns for each member constituting a household, and may be a questionnaire input obtained from a resident, NHK National Life Time Survey, ATUS (American Time Use Survey), etc. The statistical information may be input.

  The family attribute DB 111 is a database that accumulates family attributes attributed to members' occupations and roles in the household that can be defined in the action label series teacher DB 108. The family attribute DB 111 may also be configured by extracting the family attribute from the questionnaire input acquired from the resident, or may be configured by a family attribute such as NHK national life time survey. One of the member IDs defined in the family attribute DB 111 is applied to each member of the action label series teacher DB 108. For example, member ID “1” = father, “2” = mother, and “3,4” = child (elementary school student). The data in the family attribute DB 111 is a family attribute list that is used in the grouping for each family attribute in the subsequent processing, and the family attribute grouping is registered in advance. .

  2A and 2B show specific configuration examples of data stored in the action label series teacher DB 108. FIG. For each family attribute (member ID) defined in the family attribute DB 111 for each household (group), the action label series teacher DB 108 “when, how long, what did "". The names (hereinafter referred to as “behavior labels”) indicating the respective behaviors at this time are types defined in advance in a questionnaire or survey (for example, “sleep”, “meal 1”, “TV”, “bath”). Etc.). That is, FIG. 2A shows the action label series of the first day of the member ID “1, 2, 3, 4” for the group “0”, and FIG. 2B shows the member ID “1, 2” for the group “0”. , 3, 4 ″, the second day action label series. When a member takes an action to which a plurality of action labels can be applied in one time zone, a plurality of action labels can be applied such as “meal 1, cooking” or “TV, washing”.

  The action label DB 116 with a single action flag is a database for registering action labels that mainly act alone among action labels. As shown in FIGS. 2A and 2B, a single action flag (circle in FIGS. 2A and 2B) can be assigned for each action label of each member in action label units. For example, according to FIG. 2A, “sleep” and “work” with member ID “1”, “cleaning” and “cooking” with member ID “2”, “sleep” and “school” with member ID “3”, etc. Is assigned a single action flag. Moreover, you may provide in the form of "the action label of a specific member ID is independent". The single action flag is inherited as it is for the used device label when the used device series is converted. For example, if “cleaning” → “use of vacuum cleaner” is used as a single action, a flag “single” is added to the device label “vacuum cleaner” as it is, and data such as (vacuum cleaner, single) is used. Hold.

  The action label using device DB 109 is a database for registering the used device series used for the use time and the use time used when the action of each action label in the action label series teacher DB 108 is performed.

  FIG. 3 shows a specific configuration example of data stored in the action label using device DB 109. In the action label using device DB 109, for example, various sensors can be used to detect ON / OFF of lighting, heating, and the like based on the amount of current and electric power, and to detect the amount and usage time of the gas stove with the gas detection sensor at all times. The data automatically recorded using the above is registered in advance as a device series to be used. For example, there are the following four methods for ON / OFF detection. As a first method, a clamp current sensor is installed on the conductor of the existing distribution board, and ON / OFF is determined by the amount of current (≈power). A second method is to use a distribution board capable of monitoring the amount of power used, periodically measure the amount of power, and determine ON / OFF according to the amount of power, third As a method, a sensor is bridged between the power plug and the outlet, the electric energy of individual electronic devices is measured periodically, and ON / OFF is determined based on the amount of the electric energy. Some electronic devices have a communication function, and their operation information can be stored in the server upon request. Either ON / OFF information can be obtained directly from the operation information, or the electric energy of individual electronic devices can be periodically Measured, there is a method of determining the ON / OFF by the amount of the electric energy, for detecting the ON / OFF using any technique. In addition, the used device series may be registered in association with each action label in the action label series teacher DB 108.

  The action label using device conversion unit 101 matches the action label corresponding to the action label series teacher DB 108 from the data describing the correspondence between the action label defined in the action label using device DB 109 and the used device, and the device classified by family attribute The usage series DB 110 is output. At the time of this matching, the single action flag is also given to the device series used, and the assigned / non-assigned labels are distinguished as different labels. When the action label using device DB 109 itself is the series data of the usage history of each member's device, the content data is substantially the same as the family attribute-specific device usage series DB 110.

  FIG. 4 shows a processing flow of the action label using device conversion unit 101. Here, for each individual action label, the actual device usage history and the data in which the action label has been previously matched are already registered in the action label using device DB 109, and the action label type matches. Suppose that it is done.

  The action label using device conversion unit 101 selects a teacher action label series for one day from the action label series teacher DB 108 (S1301), and for each action label in the teacher action label series for one day in S1302 to S1306. Steps S1303 to S1305 are performed. The action label using device conversion unit 101 searches the action label using device DB 109 for a set of used devices that match the action label, registers them as a used device series (S1303), and the target action label is a single action (S1304), a single action flag is assigned to each device used in the action label (S1305).

  The feature use device series extraction unit 102 receives the family attribute-specific device use series DB 110 and the family attribute DB 111 as inputs. That is, device usage history series data having a specified family attribute is collected as one group for all or some households, and frequently used device usage history series is extracted as a family attribute-specific template. In the template extraction by family attribute, the presence / absence of “single action flag” is distinguished and aggregated, and the one with the highest frequency is selected.

  For the extraction method, for example, MJ Zaki, “SPADE: An Efficient Algorithm for Mining Frequent Sequences”, Machine Learning 42, 1/2, pp. 31-60 (2001), or Agrawal, R. & Srikant, R. The method proposed in “Mining sequential patterns”, 11th Intl. Conf. On Data Engineering, pp. 3-14 (1995) is used. Since a plurality of templates by family attribute are usually extracted by taking a threshold value inputted in advance or the like, all of them are registered in the template DB 112 by family attribute.

  5A, 5B, and 5C show examples of extraction from the family attribute-specific template DB 112. FIG. FIG. 5A shows a father template extracted with the member ID “1”. FIG. 5B shows a mother template extracted with the member ID “2”. FIG. 5B shows a child (primary school student) template extracted with the member ID “3, 4”.

  The family mixed use equipment state series DB 113 stores a series of equipment use histories sensed per household. FIG. 6 shows an example of a family mixed use device state series for a certain household (group ID = 101). The data collected by the family mixed use device state series DB 113 includes device use history data detected by a sensor for each household as a result of individual members or a plurality of members taking some action.

  In many cases, many devices are provided. As shown in FIG. 6, it is assumed that the device usage data in each time zone is available, but it is unclear who used it. The device usage data at that time is assumed to be ON / OFF information or data representing the operation state of each device. For example, it is possible to acquire this information using a product of FEMINITY (registered trademark), and even when data other than the above devices is generated when power consumption of a certain value or more is generated by a watt checker or the like, it is less than that value. The usage history of the device can be held by turning off the case. In addition, data such as gas and water may be used as device usage data as long as the usage amount can be measured. It is assumed that the sensor series label is defined by the action label using device DB 109.

  The mixed use equipment state series separation unit 103 uses a family attribute-specific template DB 112 from the family mixed use equipment state series data of a specific household for one day, among the data of the family use use equipment state series DB 113, to specify a specific The family attribute and the device used by the family attribute are estimated based on matching with the frequent pattern of the family attribute. Here, in the matching between the family mixed use device state series and the template, the “single action flag” information is not used, and the “use device” label is used for matching. Although it is not used as a target for matching, a single action flag is attached as it is as a label without being separated during this processing.

  The mixed use device state series separation unit 103 stores the data of the separated family attribute and the device state series data estimated as the use device as one set in the separated use device state series DB 114 with family attribute.

  7A, 7B, and 7C show an example of a method of separating the used device series for each family attribute for the used device series of each member in the household from the family mixed used device state series of FIG. Here, in the case where the family mixed use device state series is input in advance that the family attribute of the target household is generated by one of the father, mother, and child (elementary school student), the father, mother, and child Determine the match rate of the devices used in the same time period from the templates of the devices used by (elementary school students), and the device that is used by each member of the family attribute is the template with the maximum Σ (target rate) × duration A series.

  FIG. 7A shows an example in which the family mixed use device state series shown in FIG. 6 is matched with the father template (template 1) of FIG. 5A. FIG. 7B shows an example in which the family mixed use device state series shown in FIG. 6 is matched with the mother templates (templates 1 and 2) shown in FIG. 5B. FIG. 7C shows an example in which the family mixed use device state series shown in FIG. 6 is matched with the child (elementary school) template (templates 1 and 2) shown in FIG. 5C. 7A, 7 </ b> B, and 7 </ b> C, a portion that matches each template in the used device label of the family mixed used device state series is indicated by a hatched pattern. About FIG. 7B, since the matching location (12 places) of the template 1 exceeds the matching location (10 places) of the template 2, the template 1 is employ | adopted. About FIG. 7C, since the matching location (6 places) of the template 1 exceeds the matching location (3 places) of the template 2, the template 1 is employ | adopted.

  As shown in the upper part of FIG. 8, the template having a high match rate adopted as described above is set as the “selected template” of each member. The mixed use device state series separation unit 103 sets the family attribute separated by the “selected template” of each member and the device state series data estimated as the use device as a set of data (for example, the lower row in FIG. 8). The oblique line pattern portion) is stored in the separated use device state series DB 114 with family attributes.

  In this matching, the added content of the single action flag is not referred to, and the state of each used device in the mixed use device state series is compared with the state of the used device of the template.

  With reference to FIG. 9 thru | or FIG. 11, the process of the mixing use apparatus state sequence separation part 103 is demonstrated. FIG. 9 is a flowchart showing a routine for selecting a used device label conformity location and an evaluation function application order determination routine.

  The mixed use device state series separation unit 103 selects a mixed use device state series for several days (S1401), and for the mixed use device state series for each day (S1402), the family in the family attribute list For each attribute (S1403), subroutine 1 described later is executed for each family attribute template in the family attribute template set (S1404). The array output in subroutine 1 is saved for all family attribute templates (S1405).

  FIG. 10 shows subroutine 1. The mixed use device state series separation unit 103 performs, for each use device series in the use device template set of one family attribute (S1406), for each time slice of the mixed use device state series for one day per day. Then, the place where all the action labels of the template match is stored in the array (S1407). A combination of matching sequences is generated in a lattice shape (S1408). Combinations of matching sequences that do not match the template order relationship are excluded (S1409). Lattice-like combinations of locations that match the family attribute template are arranged and stored (S1410).

  FIG. 11 shows an evaluation function application routine. The mixed use equipment state series separation unit 103 selects a mixed use equipment state series for several days (S1501), and among the selected several days of mixed use equipment state series, a mixed use equipment state series for each day. (S1502), for each family attribute in the family attribute list (S1503), for each family attribute template in the family attribute template set (S1504), FIG. "Evaluation function application order determination routine".

  Regarding one application order array among all the application order arrays (S1505), when the value of the evaluation function of the application order array selected this time is larger than the value of the evaluation function of the currently applied application order array (S1506). The application order arrangement selected this time is adopted (S1507). If the highest family attribute evaluation function value is extracted (S1508), if it is not excluded from the whole family attribute search target, the process proceeds to S1502, and is excluded from the whole family attribute search target Ends (S1509).

  In addition to the examples in FIGS. 9 to 11 described above, if tf / idf is applied using the device label as a word and the total sum of the scores is used, the number of persons can be determined without inputting the family attribute of the target household. If the family is input in advance, the family attribute can be estimated and the device series used by the family attribute member can also be estimated.

  Furthermore, when an evaluation function is defined using a matching score of tf / idf or Σ (target rate) × duration as a variable, the combination is determined based on which template combination is likely. You can also.

  The group used device state label set extraction unit 104 is estimated to be a used device in the separated used device state series with family attribute, as shown in the latter stage of FIG. From the set of all member device state series data (a subset of the mixed use device state series), the set of devices that are used at a high frequency in the same time zone is defined as the “group use device state set”, and the combinations are extracted. To do. As a result, the used device set data of the devices used in common by a plurality of members is stored in the group used device state set DB 115 in a specific time zone. Note that the used device series label to which the single action flag is assigned is excluded from the extraction target when the group used device is determined. In other words, at the time of extracting the group device status label set, the device used with the single action flag is not adopted in the determination of “used in group”, but simply “behaved (different) action performed in the same time zone”. To do.

  In addition, the separated use device status series DB 114 with family attributes is taken out in household units, referenced for each time zone, and when it is determined to be used by a plurality of members, that time zone is set as a group use device status label set. . The determination result is stored in the group use device state set DB 115.

  With reference to the flowchart of FIG. 12, the process of the group use device state label set extraction unit 104 will be described. First, for a specific day of all data acquisition dates (S1511), for each unit time of all time zones (S1512), for each member of all households (S1513), vector used by each member (S1514). Then, it is determined whether or not the device is a device used with a single flag (S1515). If the device is a device used with a single flag, it is excluded from the vector (S1516), and all combinations of devices used (hereinafter referred to as group use). For the device candidate vector), the number of occurrences within the unit time is counted (S1517).

  Next, for each of the group use device candidate vectors (S1518), it is determined whether or not the group use device candidate vector occurs more than a threshold for the used combination of members (S1519). For example, the threshold can be set such that Confidence is 0.5 or more. The group-used device state label set extraction unit 104 stores the group-used device state vector that has occurred above the threshold in the group-used device state set DB 115 as a group-used device state label set that is used simultaneously (S1520).

  FIG. 13 shows a specific example of the operation of the group device status label set extraction unit 104. In FIG. 13, one day of the template selected as shown in the latter part of FIG. 8 will be described. As shown in the upper part of FIG. 13, the selected template is swept for each unit time, and in the first unit time, (individual room 2 / nightlight) and (child's room / nightlight) are assigned single action flags. Are excluded from the candidates. In the next unit time, as shown in the middle part of FIG. 13, (mother, child) is once for (oven), (mother) is once for (microwave, kitchen light 1, all light), (kitchen light) (1. All light, oven) (Mother) once, (Oven, microwave) (Mother) once, (Microwave, kitchen light 1, All light, Oven) (Mother) once Be counted.

  Such enumeration is tried for the sensor observation data as a target. Thus, as shown in the lower part of FIG. 13, as a combination of the equipment used and the member used, (Oven) (Mother, Child), (Kitchen Light 1 / All Light) (Father, Mother, Child) , (Father, Mother, Child) for (Kitchen TV), (Father, Mother, Child) for (Kitchen Light 1 / All Light, Kitchen), (Father, Child) for (Father, Mother, Child), It is stored in the group use device state set DB 115.

  The used device action label conversion unit 105 converts the template data in the family attribute-specific template 112 based on the separated used device state series data with family attribute supporting the device use label for the data in the group used device state set DB 115. Identify and convert to a plausible action label sequence. The used device action label conversion unit 105 stores the converted action label series with family attributes (including actions performed by individuals and multiple persons in the group) in the group action label set DB 119.

  First, as shown in FIG. 14, the used device action label conversion unit 105 uses the action label using device DB 109, and uses the used device state series data with family attribute of the separated used device state series DB 114 with family attribute. Convert from a series to an action label series to generate an action label series with family attributes.

  In addition, the group simultaneous action label set extraction unit 106 collects a plurality of action label sets for each time period from the action label series with family attributes, and uses a frequent label extraction method such as Priori, and a plurality of members exceed a threshold value. The group simultaneous action label set that simultaneously acts at the rate of is extracted. The extraction result is stored in the group simultaneous action label set DB 118. In other words, the group simultaneous action label set DB 118 is a DB composed of “only actions performed by a group (a plurality of persons)”, and the group action label set DB 119 is obtained from the family mixed use device state series, A DB that stores a result of estimating a series of “who performed what in a certain time zone” from the devices used regardless of a plurality of people.

  FIG. 15 shows the group simultaneous action label set extraction processing of the group simultaneous action label set extraction unit 106. Apriori is applied by setting the length of unit time and the ratio (threshold, 0.5 in this example) included in the daily unit template. Among the extracted action label series of one household unit, a set of action labels in a time period in which action labels of two or more people are extracted in the same time period is referred to as a “group simultaneous action label”. The case where there are multiple is defined as “group simultaneous action label set”.

  FIG. 16 shows a flowchart of the group simultaneous action label set extraction unit 106. The group simultaneous action label set extraction unit 106 selects a mixed action series (with family attributes) for several days (S1601), and divides all the mixed action series (with family attributes) into action label groups of one time slice. (S1602). Then, a label obtained by combining the action label and the family attribute label is generated for the action label group (with the family attribute) of one time slice (S1603). Apriori is applied to determine a subset of the action label group (with family attribute) of one time slice with high frequency (S1604).

  The group action power consumption calculation unit 107 matches the previous group action label with a device that matches the group action label, and calculates the power consumption at that time. An outline of the operation of the group action power consumption calculation unit 107 is shown in FIG.

  FIG. 17 is an example in which the amount of power consumed in the case of a plurality of actions with a time difference and a single action is compared with that in the case of simultaneous actions. By calculating such estimated statistical data of the power consumption, it is estimated what contribution to the power consumption by the group simultaneous action will be. The calculation result is stored in the power consumption DB 120 for group action according to time zone.

  For example, in FIG. 17, an action label (“watch TV in the living room”, (undefined), “watch TV in the living room”) is designated as a group simultaneous action in the family attributes (father, mother, child (elementary school student)). Indicates the case. The group action power consumption calculation unit 107 extracts the group simultaneous action time from the separated use equipment state series DB 114 with family attribute and the group use equipment state set DB 115, and the estimated power amount is the power of the action label use equipment DB 109. Calculate from usage. Further, when it is not a group simultaneous action, it is extracted from the separated use device state series DB 114 with family attribute and the group use device state set DB 115, and an estimated value of the power amount is calculated from the power use amount of the action label use device DB 109. Thereby, by comparing the estimated values of the power amounts of both, it is possible to use it for a determination such as “the group simultaneous action has a lower risk of exceeding the threshold”.

  As described above, in the above embodiment, from the behavior measurement data acquired in household units, the data obtained by the members of each attribute is estimated, divided into behavior data series for each member attribute, and Using the behavior data series, it is possible to extract behavior patterns that are frequently executed by a plurality of members simultaneously in the same time period, and to estimate the power consumption tendency at that time.

  By doing this, the power consumption when multiple members acted in the same time zone and the power consumption when the previous behavior alone in another time zone is compared, and whether or not to act as a group depending on the case It can be used as knowledge when determining whether or not.

  Furthermore, according to the above-described embodiment, a frequent pattern is obtained from the time-series data of each individual's action label, and the apparatus behavior data used when the action label is executed is collected, and the data is automatically used. Moreover, by analyzing the device use time series data in a household where a large amount of data is collected, it is possible to extract actions performed simultaneously by a plurality of members in a form that is relatively easy to implement.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 101 ... Action label use apparatus conversion part, 102 ... Feature use apparatus series extraction part, 103 ... Mixed use apparatus state series separation part, 104 ... Group use apparatus state label collection extraction part, 105 ... Use apparatus action label conversion part, 106 ... Group simultaneous action label set extraction unit, 107 ... group action power consumption calculation part, 108 ... action label series teacher DB, 109 ... action label use equipment DB, 111 ... family attribute DB, 113 ... family mixed use equipment state series DB, 116 ... Action label DB with single action flag, 110 ... Family attribute-specific equipment use series DB, 112 ... Family attribute-specific template DB, 114 ... Separate use equipment status series DB with family attributes, 115 ... Group use equipment status set DB, 118 ... group simultaneous action label set DB, 119 ... group action label set DB, 120 Time zone Group Code of power consumption DB.

Claims (5)

The behavior label sequence representing each member of behavior patterns constituting the household, and historical data and matching devices used in the practice of the action pattern, behavior label used equipment for converting the device series for behavior label used A conversion unit;
A feature use device series extraction unit that extracts a characteristic device use series as a template for each member attribute from the action label use device series;
A mixed use device state series separation unit that separates a mixed use device state sequence consisting of a set of use device state sequences of the entire household sensed in each unit time as a use device state sequence of each member using the template;
An in-household group behavior estimation apparatus comprising: a used device behavior label conversion unit that converts a used device state sequence of each member into a group behavior label set including a behavior label sequence with a family attribute.   A group simultaneous action label set for extracting a group simultaneous action label set presumed to be performed simultaneously by a plurality of members from a group action label set including an action label sequence with a family attribute output from the used device action label conversion unit The household group behavior estimation apparatus according to claim 1, further comprising an extraction unit. A group use device state label set extraction unit that extracts a group use device state label set having a high frequency of simultaneous use from the use device state series of each member;
Calculate the power consumption when multiple members act simultaneously from the group used device state label set, the group simultaneous action label set, and the history and power consumption of the used devices corresponding to the group action label set. The intra-household group behavior estimation apparatus according to claim 2, further comprising a power consumption calculation unit.   The group use device state label set extraction unit further uses the single action flag assigned to the action label acting alone to exclude the use device state series assigned with the single action flag from the extraction target. The household group behavior estimation device according to claim 3. A household group behavior estimation method executed by an information processing device,
And converting the behavior label sequence representing each member of behavior patterns constituting the household, and history matching of equipment used in the practice of the action pattern, the device series for behavior label used,
Extracting a characteristic device usage sequence from the behavior label usage device sequence as a template for each member attribute;
Separating a mixed use device state series consisting of a set of use device state sequences of the entire household sensed in each unit time as a use device state sequence of each member using the template;
A method for estimating a group behavior in a household, comprising: converting a device state series used by each member into a group behavior label set including a behavior label sequence with a family attribute.

Images