JP2021004691A - Warmth realization model generation device and method - Google Patents

Abstract

To bring an estimation result on warmth realization of a resident close to the reality.SOLUTION: A warmth realization model generation device includes: a past model generation unit 15 for generating a past model that models relations between a residence environment index of an evaluation object space and a resident's warmth realization for the evaluation object space; a season-dependent management unit 21 for obtaining a feature amount modification function indicating relations between a model feature amount that determines the shape of a function of the past model and a representative value of outdoor warmth environment based on a plurality of the past models generated over a plurality of learning data periods, and a representative value of the past outdoor warmth environment corresponding to the plurality of learning data periods; and a future model generation unit 22 for determining a future learning data period to be set for a future date in which the resident's warmth realization is desired to be estimated, calculating a model feature amount corresponding to the representative value of the outdoor warmth environment in the learning data period based on the feature amount modification function, and generating a future model modifying the latest past model based on the calculated model feature amount.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for generating a thermal sensation model for estimating a resident's thermal sensation with respect to an evaluation target space at a future date and time.

In the air conditioning control and operation management of the indoor environment, there is a demand for a technology for accurately grasping the degree of satisfaction of the resident with the indoor environment.
The resident's thermal sensation information is acquired from the resident's own declaration, the resident's behavior, images, physiological amount measurement, etc., and the acquired thermal sensation information and the index related to the resident's environment when the thermal sensation information is generated ( Hereinafter, there is a technique of constructing a learning model that learns the relationship with the living environment index) and using this learned model as an individual thermal feeling model to estimate the thermal feeling of the target resident with respect to an arbitrary environment. Residents' feelings of heat information include a measure of whole body warmth / coldness such as hot / cold, and comfort / satisfaction caused by the thermal environment. Examples of the living environment index include a physical environment index such as temperature and humidity, or a PMV (Predicted Mean Vote) which is an index using the physical environment index.

Patent Document 1 describes a method of substantially learning the relationship between the resident's feeling of warmth (declaration of hot / cold feeling in Patent Document 1) and the living environment index (indoor temperature in Patent Document 1). ..
Further, Non-Patent Document 1 describes a method of learning the relationship between the user's thermal feeling information (declaration of discomfort in Non-Patent Document 1) and the user's living environment index (PMV in Non-Patent Document 1). There is.

Residents' feelings of heat constantly change under the influence of seasonal factors (hereinafter referred to as seasonal factors) such as seasonal changes in clothing and metabolic rate, and room temperature in the bedroom at home during sleep (non-patent documents). 2). Therefore, there is a problem that an estimation error occurs when the trained model based on the past data is applied to the estimation of the thermal feeling information after the learning time, and improvement of the estimation accuracy is required.

Japanese Patent No. 2713532

Noritoshi Tanaka et al., "Study on ZEB conversion of existing offices (4th report) Research on air-conditioning control method considering individual warmth and cold feeling", Proceedings of the Conference of the Air Conditioning and Sanitary Engineering Society, Vol. 10, pp213 216, 2017/9 Hiroko Kubo et al., "Experimental Study on Seasonal Differences in Feeling of Warm and Cold", Humans and Living Environment, Vol. 1, No. 1, pp51-57, 1994

The present invention has been made to solve the above problems, and an object of the present invention is to provide a thermal sensation model generation device and a method capable of bringing the estimation result of the resident's thermal sensation closer to the actual situation.

The thermal sensation model generator of the present invention has information on the update timing and learning data period of the first model that models the relationship between the living environment index of the evaluation target space and the resident's thermal sensation with respect to the evaluation target space. The first model is based on the data period management unit configured to store the data, the living environment index in the learning data period, and the resident's thermal feeling information at the update timing. A first model generator configured to generate, and a past outdoor thermal environment corresponding to each of the plurality of first models and the plurality of training data periods generated by the plurality of training data periods. A seasonally dependent management unit configured to obtain a characteristic quantity correction function indicating the relationship between the model characteristic quantity that determines the shape of the function of the first model and the representative value of the outdoor thermal environment based on the representative value of. Then, the future learning data period to be set for the future date and time for which the resident's feeling of heat is to be estimated is determined, and the model characteristic quantity corresponding to the representative value of the outdoor thermal environment in this learning data period is calculated. It is provided with a second model generation unit configured to generate a second model obtained by calculating based on a characteristic amount correction function and modifying the latest first model based on the calculated model characteristic amount. It is characterized by that.

Further, in one configuration example of the thermal feeling model generation device of the present invention, the data period management unit stores information on the overlap period in addition to the information on the update timing and the information on the learning data period, and the plurality of data periods are stored. The learning data period of the above is characterized in that the learning data periods before and after the continuation are set so as to overlap by the width of the overlapping period.
Further, one configuration example of the thermal experience model generation device of the present invention further includes an outdoor thermal environment information management unit configured to store past outdoor thermal environment information of the building including the evaluation target space. The seasonally dependent management unit acquires the past outdoor thermal environment information corresponding to each of the plurality of learning data periods from the outdoor thermal environment information management unit, and the past outdoor corresponding to each of the plurality of learning data periods. The representative value of the thermal environment is determined, the characteristic quantity correction function is obtained, and the second model generation unit manages the past outdoor thermal environment information corresponding to the future learning data period to the outdoor thermal environment information management. It is characterized in that the representative value of the outdoor thermal environment corresponding to the future learning data period is determined by acquiring from the department.
Further, in one configuration example of the thermal feeling model generation device of the present invention, the second model generation unit has a model characteristic amount calculated based on the characteristic amount correction function and the latest model characteristic amount of the first model. It is characterized in that the latest first model is modified based on the above.
Further, in one configuration example of the thermal sensation model generator of the present invention, the living environment index is PMV, and the thermal sensation of the resident is the ratio of the resident who declares dissatisfaction with the thermal environment.

Further, in the method for generating a thermal sensation model of the present invention, information on the update timing of the first model that models the relationship between the living environment index of the evaluation target space and the resident's thermal sensation with respect to the evaluation target space and the learning data period. With reference to the data period management unit that stores the information of the above, when the update timing is reached, the first model is created based on the living environment index in the learning data period and the thermal feeling information of the resident. Based on the first step to generate, the plurality of first models generated by the plurality of training data periods, and the representative values of the past outdoor thermal environment corresponding to each of the plurality of training data periods. The second step of finding the characteristic quantity correction function showing the relationship between the model characteristic quantity that determines the shape of the function of the first model and the representative value of the outdoor thermal environment, and the future in which the resident's thermal feeling is to be estimated. The future training data period to be set for the date and time was determined, and the model characteristic quantity corresponding to the representative value of the outdoor thermal environment in this training data period was calculated and calculated based on the characteristic quantity correction function. It is characterized by including a third step of generating a second model obtained by modifying the latest first model based on a model characteristic quantity.

Further, in one configuration example of the thermal feeling model generation method of the present invention, the data period management unit stores information on the overlap period in addition to the information on the update timing and the information on the learning data period, and the plurality of data periods are stored. The learning data period of the above is characterized in that the learning data periods before and after the continuation are set so as to overlap by the width of the overlapping period.
Further, in one configuration example of the thermal experience model generation method of the present invention, the second step refers to the outdoor thermal environment information management unit that stores the past outdoor thermal environment information of the building including the evaluation target space. , The past outdoor thermal environment information corresponding to each of the plurality of learning data periods is acquired from the outdoor thermal environment information management unit, and a representative value of the past outdoor thermal environment corresponding to each of the plurality of learning data periods. Including the step of determining the characteristic quantity correction function and obtaining the characteristic quantity correction function, the third step acquires the past outdoor thermal environment information corresponding to the future learning data period from the outdoor thermal environment information management unit. The present invention is characterized by including a step of determining a representative value of the outdoor thermal environment corresponding to the future learning data period.
Further, in one configuration example of the method for generating a thermal feeling model of the present invention, the third step is divided into a model characteristic amount calculated based on the characteristic amount correction function and the latest model characteristic amount of the first model. Based on this, it is characterized by including a step of modifying the latest first model.

According to the present invention, the data period management unit, the first model generation unit, the season-dependent management unit, and the second model generation unit are provided, and the first model is based on the outdoor thermal environment state in which the seasonal transition is reflected. By modifying the above to generate a second model, it is possible to improve the estimation error due to the seasonal transition element. As a result, in the present invention, the estimation result of the resident's feeling of heat can be brought closer to the actual situation.

Further, in the present invention, in addition to the update timing information and the learning data period information, the overlap period information is stored in the data period management unit, and a plurality of consecutive learning data periods are limited to the width of the overlap period. By setting them to overlap, the reliability of the model can be further improved.

FIG. 1 is a diagram showing an example of an air conditioning control system according to a first embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a thermal sensation model generator according to the first embodiment of the present invention. FIG. 3 is a flowchart illustrating the operation of the living environment information holding unit, the heat sensation information holding unit, and the data integration unit of the thermal sensation model generator according to the first embodiment of the present invention. FIG. 4 is a diagram showing an example of the warm / cold feeling declaration information, the declaration date / time, and the filer ID in the first embodiment of the present invention. FIG. 5 is a diagram showing an example of integrated information in the first embodiment of the present invention. FIG. 6 is a flowchart illustrating the operation of the past model generation unit of the thermal experience model generation device according to the first embodiment of the present invention. FIG. 7 is a diagram showing the relationship between PMV, which is a living environment index, and the number of filers who report “hot”. FIG. 8 is a flowchart illustrating a method of generating modeling information according to the first embodiment of the present invention. FIG. 9 is a diagram showing the relationship between PMV and PPD. FIG. 10 is a diagram showing an example of the dissatisfaction function. FIG. 11 is a flowchart illustrating the operation of the season-dependent management unit of the thermal sensation model generator according to the first embodiment of the present invention. FIG. 12 is a flowchart illustrating the operation of determining the characteristic quantity correction function of the season-dependent management unit of the thermal feeling model generation device according to the first embodiment of the present invention. FIG. 13 is a diagram illustrating a model characteristic quantity in the first embodiment of the present invention. FIG. 14 is a diagram illustrating a method for determining a characteristic quantity correction function in the first embodiment of the present invention. FIG. 15 is a diagram illustrating another model characteristic quantity in the first embodiment of the present invention. FIG. 16 is a flowchart illustrating the operation of the future model generation unit of the thermal feeling model generation device according to the first embodiment of the present invention. FIG. 17 is a diagram illustrating a learning data period of the first and second embodiments of the present invention. FIG. 18 is a diagram showing an example of a period average transition value of the outside air temperature. FIG. 19 is a block diagram showing a configuration example of a computer that realizes the thermal sensation model generator according to the first and second embodiments of the present invention.

[Principle of invention]
When there is a seasonal transition, it is necessary to respond to the change in the actual state of thermal sensation when the learning data is collected. Therefore, the inventor focused on modifying the trained individual thermal experience model based on past data (hereinafter referred to as the past model) based on the outdoor thermal environment condition reflecting the seasonal transition to make it a future model. It is possible to reflect the influence of seasonal fluctuations in the model by modifying the past model in the outdoor thermal environment condition that reflects the seasonal transition.

Outdoor thermal environment conditions that reflect seasonal changes include outdoor outside air temperature. For example, a future model is generated from a trained past model by using the average outside temperature during the training data period when the past model is generated and the future normal outside temperature or outside temperature forecast value for which the resident's feeling of heat is to be estimated. To do.

Furthermore, the inventor noted that the reliability of past and future models cannot be maintained when using data for periods that include short-term fluctuations such as 2 days and 3 days. In order to take into account the annual fluctuation of seasonal transition, it is necessary to secure data with a certain period width corresponding to the cycle of this fluctuation for model training. A suitable period width is about one month, but in order to maintain the continuity of seasonal transitions, an overlap method is adopted in which a reliable period (such as two weeks) overlaps between the previous period and the later period. I came up with the idea that is effective.

[First Example]
Hereinafter, examples of the present invention will be described with reference to the drawings. In the first embodiment of the present invention, a basic method for generating a future model from a past model will be described.
In this example, a model for each month is generated on a monthly basis, and a past model is based on the data period month's normal outside temperature of the past model and the future model data period month's normal outside temperature including the future to be predicted. The model parameter of is modified by the characteristic quantity correction function, and the past model is modified by using the characteristic quantity correction function to make it a future model.

In this embodiment, for the sake of simplicity, as shown in FIG. 1, an example of a resident declaration type air conditioning control system in which two residents are present in one target air conditioning zone will be described. Regarding the resident-declared air-conditioning control system, for example, the literature "Yasuhito Omagari et al.," Demonstration test of air-conditioning system for hot and cold feeling declaration ", Proceedings of the Society of Air Conditioning and Sanitary Engineering, Vol. 3, pp41-44, 2016 / 9 ”is disclosed.

In FIG. 1, 100 is a resident, 101 is an evaluation target space (resident's seating space), 102 is an air conditioning controller (controller) that receives a hot / cold feeling report, and 103 is a temperature for measuring the room temperature of the evaluation target space 101. The sensor, 104 is a humidity sensor that measures the humidity of the evaluation target space 101, 105 is an indoor unit, and 106 is an outdoor unit. The air conditioner controller 102 is an air conditioner (indoor unit 105 and outdoor unit 106) so that the room temperature measured by the temperature sensor 103 matches the room temperature set value and the humidity measured by the humidity sensor 104 matches the humidity set value. To control.

FIG. 2 is a block diagram showing a configuration of a thermal experience model generator of this embodiment. The thermal feeling model generation device is composed of a past model generation unit 1 and a future model generation unit 2.

The past model generation unit 1 is the data collection unit 10, and information and learning of the update timing of the individual thermal sensation model that models the relationship between the living environment index of the evaluation target space 101 and the resident's thermal sensation with respect to the evaluation target space 101. The past model (first model) is created based on the data period management unit 14 that stores the data period information, and the living environment index and the resident's thermal feeling information in the learning data period when the update timing comes. It includes a past model generation unit 15 (first model generation unit) to be generated, and a past model storage management unit 16 that stores the past model generated by the past model generation unit 15.

The data collection unit 10 includes a living environment information holding unit 11 that calculates a living environment index from the thermal environment information, a thermal feeling information holding unit 12 that acquires the thermal feeling information declared by the resident of the evaluation target space, and a thermal feeling. It is composed of a data integration unit 13 that extracts the living environment index calculated from the thermal environment information of the measurement date and time corresponding to the generation date and time of the information and generates integrated information that integrates the thermal feeling information and the living environment index. ..

The future model generation unit 2 includes an outdoor thermal environment information management unit 20 that stores past outdoor thermal environment information of the building including the evaluation target space, and a plurality of past models and a plurality of past models generated by a plurality of learning data periods. Based on the representative values of the past outdoor thermal environment corresponding to each of the training data periods, the characteristic quantity correction function showing the relationship between the model characteristic quantity that determines the shape of the function of the past model and the representative value of the outdoor thermal environment Determine the future learning data period to be set for the desired season-dependent management unit 21 and the future date and time for which the resident's feeling of heat is to be estimated, and correspond to the representative value of the outdoor thermal environment in this learning data period. Future model generator 22 (second model) that calculates the model characteristic quantity based on the characteristic quantity correction function and generates a future model (second model) in which the latest past model is modified based on the calculated model characteristic quantity. It has a model generator).

FIG. 3 is a flowchart illustrating the operation of the living environment information holding unit 11, the thermal feeling information holding unit 12, and the data integration unit 13.
The living environment information holding unit 11 obtains thermal environment information (thermal environment measurement values such as temperature measurement value and humidity measurement value) from, for example, an environment measurement device (temperature sensor 103, humidity sensor 104, etc.) installed in the evaluation target space 101. Receive and calculate the living environment index. Then, the living environment information holding unit 11 stores the measurement date and time of the thermal environment measurement value and the living environment index as living environment information (measurement date and time, living environment index) in association with each other (step S100 in FIG. 3).

In the living environment information holding unit 11, the calculation formula and parameter values necessary for calculating the living environment index, information on the type of thermal environment measurement value used for the calculation, and the like are provided to the solution provider such as indoor environment evaluation / control and facility management. It is preset by the person.

The living environment information holding unit 11 transmits the living environment information (measurement date and time, living environment index) to the data integration unit 13 in response to the information transmission request from the data integration unit 13 (step S101 in FIG. 3).

The living environment index is a thermal environmental index of the living environment, and is a physical environment index such as air temperature and humidity generally measured in an air-conditioned environment, or a general index calculated using the physical environment index (for example). Operating temperature, SET ^* (Standard new Effective Temperature), PMV, etc.). In this embodiment, the living environment index is PMV, which is internationally standardized, and the six elements normally required for PMV calculation: air temperature Ta [° C], humidity H [%], average radiation temperature Tr [° C], and airflow. The speed v [m / s], the amount of metabolism M [met], and the amount of clothing Icl [clo] are set to the following values, respectively. That is, the air temperature Ta [° C.] and the humidity H [%] are the measured values of the air temperature and the measured humidity of the target building, the average radiation temperature Tr is set to be equal to the air temperature Ta, and the airflow velocity v is 0.1. An example of obtaining PMV will be described with [m / s], a metabolic amount M of 1.0 [met], and a clothing amount of Icl of 0.8 [cl]. Other than these, it is assumed that the information required for PMV calculation is set by referring to the international standard and the like together with the PMV calculation formula.

Here, a method of calculating PMV will be described. Since the PMV calculation method is internationally standardized by ISO-7730, the calculation method of this standard may be followed. PMV is obtained by using the following formulas (1) to (3). The calculation program is also published in ANSI / ASHRAE Standard 55-2010 and the like.
PMV = Q (M) x L ... (1)
Q (M) = 0.303exp ^-0.036M + 0.028 ... (2)
L = M- (C + R + Ed + Es)-(Cre + Ere) ... (3)

M is the amount of metabolism [met], L is the heat load of the human body L [W / m ² ], C is the amount of convection heat loss [W / m ² ], R is the amount of radiant heat loss [W / m ² ], and Ed is Insensitive steam amount [W / m ² ], Es is evaporation heat loss amount due to sweating [W / m ² ], Cre is manifest heat loss amount due to breathing [W / m ² ], Ere is latent heat loss amount due to breathing [W] / M ² ].

The following equations (4) to (6) are convergently calculated to calculate the convection heat loss amount C and the radiant heat loss amount R. tcl is the clothing outer surface temperature [° C.].
C = fcl × hc × (tcl-Ta) ・ ・ ・ (4)
R = 3.96 × 10 ^-8 × fcl × {(tcl + 273.15) ⁴
-(Tr + 273.15) ⁴ } ... (5)
tcl = ts-0.155 x Icl x (C + R) ... (6)

In formulas (4) to (6), Icl is the amount of clothing [clo]. fcl is a clothing area increase coefficient, and when Icl <0.78, fcl = 1 + 1.29 × Icl, and when Icl ≧ 0.78, fcl = 1.05 + 0.645 × Icl. Tr is the average radiation temperature [° C.]. ts is the average skin temperature [° C.] and is calculated by the following equation. W is the mechanical work load [W / m ² ] (usually 0).
ts = 35.7-0.028 × (MW) ・ ・ ・ (7)

Further, hc is the convective heat transfer coefficient [W / (m ² · ° C.)] of the human body, which is the larger of 2.38 × | tcl-ta | ^0.25 or 12.1 × v ^0.5 . v is the airflow velocity [m / s].

The insensitive steaming amount Ed is obtained by the following equation.
Ed = 3.05 × (5.73-0.007 × M-pa) ・ ・ ・ (8)
The amount of heat of vaporization Es due to sweating is obtained by the following equation.
Es = 0.42 × (M-58.15) ・ ・ ・ (9)

The amount of sensible heat loss Cre due to respiration is obtained by the following equation.
Cre = 0.0014 × M × (34-Ta) ・ ・ ・ (10)
The latent heat loss amount Ere due to respiration is obtained by the following equation. pa is the vapor pressure.
Ere = 0.0173 × M × (5.87-pa) ・ ・ ・ (11)

From the above, the living environment information holding unit 11 can calculate PMV as a living environment index. In this embodiment, it is assumed that the thermal environment measurement value measured in a 1-minute cycle is transmitted to the living environment information holding unit 11, and the living environment information holding unit 11 calculates the PMV in a 1-minute cycle. If the thermal environment measurement values are periodically accumulated in the air conditioning control device or the central monitoring system that manages the building, the thermal environment measurement values may be acquired from these air conditioning control devices or the central monitoring system.

On the other hand, the heat feeling information holding unit 12 receives the heat feeling information of the resident of the evaluation target space 101, and associates the heat feeling information with the information of the resident who declared the heat feeling information and the generation date and time of the heat feeling information. Store it (step S102 in FIG. 3).

The thermal feeling information holding unit 12 transmits the thermal feeling information, the occurrence date / time information, and the resident information to the data integrating unit 13 in response to the information transmission request from the data integration unit 13 (step S103 in FIG. 3).

The thermal sensation information is an amount indicating the resident's feeling corresponding to the living environment index, and indicates information such as a scale of whole body warming sensation such as hot / cold and comfort / satisfaction caused by the thermal environment. .. Indicators such as the number of dissatisfied persons in the target space and the dissatisfied person rate calculated from the number of dissatisfied persons are also included in the thermal feeling information. As an input terminal for the heat feeling information, a smartphone, a PC (personal computer), or the like may be appropriately used.

In this embodiment, the hot / cold feeling report information (hot / cold) from the resident is used as the hot / cold feeling information, and the resident freely makes a hot / cold feeling report through the input terminal. FIG. 4 shows an example of the warm / cold feeling declaration information, the declaration date / time (occurrence date / time information), and the reporter ID (resident information). In the example of FIG. 4, hot is indicated by "hot" and cold is indicated by "cold".

The reporter ID may be input by the resident together with the heat / cold feeling report and transmitted to the heat / cold feeling information holding unit 12, or when the individual resident uses the input terminal, the terminal ID is used as the filer ID. , May be transmitted to the thermal feeling information holding unit 12.
Further, the date and time of the declaration may be transmitted by the input terminal to the heat sensation information holding unit 12, or may be the date and time when the heat sensation information holding unit 12 receives the heat sensation report information.

Next, when the data integration unit 13 receives the living environment index from the living environment information holding unit 11 and receives the heat feeling information from the heat feeling information holding unit 12, the data integration unit 13 resides on the measurement date and time corresponding to the generation date and time of the heat feeling information. The environmental index is extracted, and integrated information (occurrence date / time information, resident ID, thermal feeling information, living environment index) that integrates the thermal feeling information and the living environment index is stored (FIG. 3, step S104).

The data integration unit 13 transmits the integrated information to the past model generation unit 15 in response to the information transmission request from the past model generation unit 15 (step S105 in FIG. 3).

FIG. 5 shows an example of integrated information consisting of heat feeling information generation date / time information (declaration date / time information), resident ID (reporter ID), heat feeling information, and living environment index (PMV).
When extracting the living environment index of the measurement date and time corresponding to the occurrence date and time of the thermal feeling information, the resolution of the generation date and time (declaration date and time) of the thermal actual feeling information is, for example, 1 minute unit, and the resolution of the measurement date and time of the thermal environment measurement value For example, the time interval between the generation of the thermal feeling information and the measurement of the thermal environment measurement value may be different, such as in units of 10 minutes. In this case, the solution provider or the equipment manager may determine the extraction rule in advance.

For example, if the resolution of the date and time of occurrence of the heat sensation information (declaration date and time) is in units of 1 minute and the measured value of the thermal environment is measured every 10 minutes, the latest measurement date and time of the heat sensation information or the heat sensation information It is sufficient to determine the extraction rule that the date and time when the occurrence date and time is advanced in units of 10 minutes is set as the measurement date and time corresponding to the occurrence date and time of the thermal feeling information.

When the thermal environment measurement value is measured every 10 minutes, such as 14:50, 15:00, 15:10, etc., the measurement corresponds to the occurrence date and time of the thermal feeling information at 15:03. The date and time will be 15:10, which is the latest 15:00 or 15:03, which is advanced by 10 minutes.

The living environment information holding unit 11, the thermal feeling information holding unit 12, and the data integration unit 13 continuously execute the process of FIG. 3, and the integrated information is accumulated in the data integration unit 13.

Next, in the data period management unit 14, the data period management information for determining the update timing and the learning data period of the individual thermal feeling model (past model) is provided to the solution provider or equipment manager such as indoor environment evaluation / control. It is preset by.
The data period management unit 14 transmits data period management information in response to an information transmission request from the past model generation unit 15 or the future model generation unit 22.

In this embodiment, the update timing information of the individual thermal feeling model and the learning data period width information are used as data period management information. Specifically, the update timing of the individual thermal feeling model is set to 0:00 on the first day of every month (X month 1, X is an arbitrary month), and the learning data period width is updated from the previous update timing to this time. One month until the timing. That is, since the update timing of the individual thermal feeling model is the first day of every month, the start date and time of the learning data period for updating this individual thermal feeling model is 0:00 on the first day of the previous month, and the end date and time of the learning data period. Will be 23:59 on the last day of the previous month.

Since it is only necessary to determine the update timing and training data period of the individual thermal feeling model from the data period management information, the update start date and time and the period width of the learning data (every 14 days, every 168 hours, etc.) that goes back from the update start date and time Of course, may be set as data period management information.

FIG. 6 is a flowchart illustrating the operation of the past model generation unit 15. When the past model generation unit 15 recognizes that the update timing of the individual thermal feeling model has come based on the data period management information acquired from the data period management unit 14 (YES in step S200 of FIG. 6), it corresponds to the current model update. The start date and time and the end date and time of the learning data period to be performed are determined (FIG. 6, step S201).

The past model generation unit 15 acquires the integrated information of the determined learning data period from the data integration unit 13 (step S202 in FIG. 6), and uses the acquired integrated information to input / output for identifying the individual thermal feeling model. Data (modeling information) is generated (FIG. 6, step S203). The past model generation unit 15 uses the generated modeling information to identify the individual thermal sensation model, and updates the individual thermal sensation model as the latest past model (FIG. 6, step S204).

Then, the past model generation unit 15 transmits the modeling result of the latest past model and the information of the learning data period used for modeling to the past model storage management unit 16 (step S205 in FIG. 6).
The past model generation unit 15 performs the process of FIG. 6 at each model update timing.

In this embodiment, as described above, the update timing of the individual thermal feeling model is set to 0:00 on the first day of every month. For example, when the individual thermal feeling model is updated at 0:00 on March 1, 2018, the learning data period is from 0:00 on February 1, 2018 to 23:59 on February 28, 2018. If the individual thermal experience model is updated at 0:00 on August 1, 2019, the learning data period will be from 0:00 on July 1, 2018 to 23:59 on July 31, 2018.

Further, in this embodiment, the living environment index (PMV) and the reporter rate PPVh of "hot" are used as modeling information, and the function of the predicted dissatisfaction rate PPD (Predicted Percentage Dissatisfied) is used as the basic function of the individual thermal feeling model. To do. PPD has also been internationally standardized, and ISO-7730 shows an arithmetic expression using PMV.

Next, a method of generating modeling information in step S203 will be described. As described above, the past model generation unit 15 generates modeling information from the integrated information.
The rate of filers who say "hot" in any PMV PPVh is calculated by the following equation, where N is the number of reporters who say "hot" in any PMV and the number of residents in the evaluation target space is N_all.
PPVh = N / N_all ... (12)

The number of residents N_all in the evaluation target space is preset by the solution provider for indoor environment evaluation / control and the facility manager.
Here, the number of filers N in the formula (12) may be the number of filers calculated in consideration of the physiological validity for the feeling of warmth and cold. This calculation method is based on the principle that a person who declares "hot" in any PMV feels hotter in the PMV on the hotter side than the environment.

Specifically, each filer latches the PMV that was first declared when the PMV is low (the heat is weak) at the latch start point PMV start (m) (m is a positive integer, m = in this embodiment. In the PMV region larger than the latch start point PMV start (m), the number of people is calculated while maintaining the state of occurrence of the declaration of "hot". According to this calculation method, as shown in FIG. 7, the number of reporters who say "hot" increases as the environment becomes hotter, and the physiological validity becomes higher. Hereinafter, this method will be described as the declaration latch method, and an example of generating modeling information by this declaration latch method will be described.

FIG. 8 shows a flowchart for generating modeling information by calculating the filer rate PPVh from the number of filers calculated by using the report latch method. The past model generation unit 15 extracts only the information of the declaration of "hot" from the integrated information acquired in step S202 (step S300 of FIG. 8), and the number of filers with different filer IDs included in the extracted information ("" The number of residents who declared "hot") N_vote is requested (Fig. 8, step S301). Assuming that the data example shown in FIG. 5 is the integrated information of the learning data period, the number of filers who have declared "hot" is N_vote = 4.

Next, the past model generation unit 15 initializes the variable m (m is a positive integer less than or equal to N_vote) 1 for counting the filers (step S302 in FIG. 8).
The past model generation unit 15 obtains the minimum value of the PMV when the declaration of "hot" is made for the m-th filer as the latch start point PMV start (m) (step S303 in FIG. 8).

The past model generation unit 15 determines whether m = N_vote, that is, whether or not the processing of step S303 has been completed for all the filers who have declared “hot” (step S304 in FIG. 8), and if the variable m is less than N_vote, Assuming that the processing is not completed, the variable m is incremented by 1 (step S305 in FIG. 8). In this way, the process of step S303 is performed for each filer who declares "hot".

The past model generation unit 15 has m = N_vote, and when the processing of step S303 is completed for all the filers who have declared “hot” (YES in step S304), any PMV and the corresponding “hot” Modeling information consisting of the filer rate PPVh is generated (FIG. 8 step S306).

The state of presence / absence of declaration of the filer P (m) for any PMV When V_latch (m) is PMV <PMVstart (m), V_latch (m) = 0 (no declaration), PMVstart (m) ≤ PMV , V_latch (m) = 1 (with declaration).
At this time, the reporter rate PPVh of "hot" for any PMV is obtained by the following equation.

As a result, the calculation of the reporter rate PPVh of "hot" is completed, and the generation of modeling information including the PMV of the learning data period and the reporter rate PPVh of "hot" in the learning data period is completed. In the case of the data example of FIG. 5, the vertical axis of FIG. 7 is divided by the number of filers who say "hot" to obtain the rate of filers who say "hot", which is the modeling information. The modeling information may be generated at regular intervals of PMV, or may be data only at points where PPVh increases.

Next, the method of identifying the individual thermal feeling model in step S204 will be described. As described above, in this embodiment, the predicted dissatisfaction rate PPD is used as the basic function of the individual thermal experience model. PPD is represented by the following relational expression as a function of PMV.
PPD = F (PMV) = 100-95 x exp (-0.03353 x PMV ⁴⁾
−0.2179 × PMV ² ) [%] ・ ・ ・ (14)

The relationship between PMV and PPD is as shown in FIG. The individual thermal feeling model may be defined as a V-shaped distribution function as shown in FIG. 9 in which PMV and PPD are associated with each other. Alternatively, it is also possible to separate the degree of dissatisfaction on the hot side and the cold side and define the dissatisfaction function (individual thermal sensation model).

FIG. 10 is a diagram showing an example of a dissatisfaction function disclosed in Japanese Patent Application Laid-Open No. 2015-4480. The dissatisfaction function shown in 110 of FIG. 10 may be used to obtain the dissatisfaction of "hot", and the dissatisfaction function shown in 111 of FIG. 10 may be used to obtain the dissatisfaction of "cold". do it. Examples of the function as shown in FIG. 10 include a sigmoid function, and the function of the individual thermal feeling model may be such a general model.

In this embodiment, the basic function of the equation (14) is modified by reflecting the resident's feeling of heat, and the model of the following formula (15) is used as an individual feeling of heat model (dissatisfaction model). To do. The shape of the individual thermal feeling model of the equation (15) is the same as the function shown by 110 in FIG. In the formula (15), the predicted dissatisfaction rate PPD of the formula (14) corresponding to the degree of dissatisfaction is set to the reporter rate PPVh of "hot" reflecting the feeling of heat of the resident.

In equation (15), a, b, c, d, and e are parameters of the model to be searched. Equation (16) is the one in which a = b (the minimum value of the reporter rate PPVh of "hot" is 0) in equation (15), and when this is used as the model function of the individual thermal feeling model, the search target. The parameters of the model are a, c, d, and e.

The model function of the individual thermal feeling model such as equation (15) or equation (16), the parameters of the model to be searched, and the optimization method to be used are determined by the solution provider or equipment manager such as indoor environment evaluation / control. , Preset.

The past model generation unit 15 uses the modeling information generated in step S203 to search the model function of the individual thermal feeling model preset by the solution provider or the equipment manager. Search the parameters to identify the individual thermal sensation model. As the optimization method used when searching for parameters, a general-purpose method such as the least squares method or the simplex method may be used.
With the above, the update of the past model is completed.

The past model storage management unit 16 stores the modeling result (parameter value of the searched model) of the past model received from the past model generation unit 15 and the training data period information, and stores the model information from the season-dependent management unit 21. In response to the transmission request, the stored information is transmitted to the seasonal dependence management unit 21.

The outdoor thermal environment information management unit 20 stores the past outdoor thermal environment information and the future outdoor thermal environment information of the building including the evaluation target space 101. The past outdoor thermal environment information consists of outdoor thermal environment status information and information on the date and time when the outdoor thermal environment status information was acquired. The future outdoor thermal environment information is information necessary for estimating the future outdoor thermal environment, and includes information on the outdoor thermal environment state information and information on the date and time when the appearance of the outdoor thermal environment state information is estimated.

As outdoor thermal environment state information that reflects seasonal changes, there are indicators such as outdoor air temperature and humidity, an index such as a discomfort index that integrates temperature and humidity, and WBGT (wet-bulb globe temperature). Further, the past outdoor thermal environment information may be a statistic such as normal year data (monthly average temperature, daily average temperature, etc.), or may be a measured value by a temperature sensor or the like installed outdoors. The future outdoor thermal environment information is essentially the past outdoor thermal environment information corresponding to the future learning data period, and is similar to the past outdoor thermal environment information as the past outdoor thermal environment information is reproduced in the future. It may be used as an estimated value of outdoor thermal environment information estimated based on a weather forecast or the like.

The method of acquiring outdoor thermal environment information is preset by the solution provider such as indoor environment evaluation / control and the equipment manager. When data available in advance such as monthly average temperature is used as the outdoor thermal environment state information, these data may be set in the outdoor thermal environment information management unit 20 in advance.
In this embodiment, the monthly average outside temperature of each month of the normal year is used as past and future outdoor thermal environmental condition information.

Next, the seasonal dependence management unit 21 determines the characteristic quantity correction function Gpm in response to the request from the future model generation unit 22, and the characteristic quantity correction function Gpm and the past model stored in the past model storage management unit 16 The modeling result (parameter value of the searched model) is transmitted to the future model generation unit 22.

FIG. 11 is a flowchart illustrating the operation of the seasonal dependence management unit 21. The season-dependent management unit 21 used the modeling results of N_ref times (N_ref is a preset number of reference models) from the latest modeling result to the modeling result of (N_ref-1) times before, and modeling of these N_ref times. Information on the training data period is acquired from the past model storage management unit 16 (step S400 in FIG. 11).

Subsequently, the seasonally dependent management unit 21 acquires the past outdoor thermal environment information corresponding to each of the learning data periods used for the modeling of N_ref times from the outdoor thermal environment information management unit 20 (FIG. 11, step S401).

Then, the seasonal dependence management unit 21 calculates the past outdoor thermal environment representative value Eot corresponding to the learning data period used for modeling N_ref times for each learning data period based on the acquired past outdoor thermal environment information. (FIG. 11 step S402). For example, the outdoor thermal environment state information acquired from the past outdoor thermal environment information management unit 20 is the outside air temperature measured every hour, and the learning data period is from 0:00 on August 1, 2018 to August 10, 2018. At 23:59, the average value (average outside air temperature) of the outside air temperature during this learning data period may be calculated and used as the past outdoor thermal environment representative value Eot.

However, as described above, when the monthly average outside temperature of each month of the normal year is used as the past outdoor thermal environment state information, the seasonal dependence management unit 21 keeps the past monthly average outside temperature of the month including the learning data period as it is. The representative value of the outdoor thermal environment of Eot may be used. In addition, when the training data period is set over a plurality of months, the average value of the monthly average outside temperature of these multiple months and the weighted average value according to the number of days can be set as the past outdoor thermal environment representative value Eot. Good.

Next, the seasonal dependence management unit 21 obtains a characteristic quantity correction function Gpm showing the relationship between the past outdoor thermal environment representative value Eot and the model characteristic quantity SZ of the past model (FIG. 11, step S403). The model characteristic quantity SZ is one of the quantities that determines the shape of the function of the past model. Once the characteristic quantity correction function Gpm is determined, the model characteristic quantity SZ corresponding to an arbitrary outdoor thermal environment representative value Eot can be estimated.

FIG. 12 is a flowchart illustrating the operation of determining the characteristic quantity correction function of the seasonal dependence management unit 21. In the following description, i is a variable for counting past models, and is an integer of 1 to N_ref. The smaller the variable i, the newer the past model.

Further, as shown in FIG. 13, the model characteristic quantity SZ is set to the saturation value of the reporter rate PPVh of "hot", and the characteristic quantity correction function Gpm is the past outdoor thermal environment representative value Eot and the model characteristic quantity SZ. Let it be a function of linear approximation. That is, if an arbitrary outdoor thermal environment representative value Eot is determined, the estimated value of the model characteristic amount SZ corresponding to this can be calculated by the characteristic amount correction function Gpm. When the individual thermal feeling model is set in the equations (15) and (16), the saturation value of the filer rate PPVh corresponds to a of the model search parameters a, b, c, d, and e.

The seasonally dependent management unit 21 sets the modeling result Sp (i) of N_ref times in step S400, that is, the search parameter value searched for the i-th past model (in this embodiment, the searched parameter a (in this embodiment). i), b (i), c (i), d (i), e (i) set) and the information of the training data period LP (i) used for modeling these N_ref times are stored in the past model. After acquiring from the part 16, the variable i is initialized to 1 (step S500 in FIG. 12).

The seasonal dependence management unit 21 acquires the past outdoor thermal environment representative value Eot (i) corresponding to the learning data period LP (i) used for the i-th modeling from the latest from the values calculated in step S402 ( FIG. 12 step S501).

The season-dependent management unit 21 determines whether i = N_ref, that is, whether or not the processing of step S501 has been completed for all N_ref training data period LPs (i) (FIG. 12 step S502), and when the variable i is less than N_ref. The variable i is incremented by 1 (FIG. 12, step S503), assuming that the processing has not been completed. In this way, the process of step S501 is performed for each learning data period LP (i).

The season-dependent management unit 21 sets i = N_ref, and when the processing of step S501 is completed for all the training data period LPs (i) of N_ref (YES in step S502), the function and parameter search of the individual thermal feeling model Find a function that linearly approximates the relationship between the model characteristic quantity SZ (i) determined from the modeling result Sp (i), which is a set of values, and the outdoor thermal environment representative value Eot (i), and use this function as the characteristic quantity correction function. Let it be Gpm (step S504 in FIG. 12). Here, in this embodiment in which the model characteristic quantity SZ is the saturation value of the reporter rate PPVh of “hot”, SZ (i) = a (i).

14 (A) and 14 (B) are diagrams illustrating a method for determining the characteristic quantity correction function Gpm. In the examples of FIGS. 14 (A) and 14 (B), N_ref = 3, the function of the latest past model is Mp (1), the function of the past model one time before the latest is Mp (2), and from the latest. The function of the past model two times ago is Mp (3).

For example, the training data period LP (1) used when modeling the latest past model is in September, and the training data period LP (2) used when modeling the previous past model is in August and twice before. Assuming that the learning data period LP (3) used in modeling the past model is July, the past outdoor thermal environment representative values Eot (1), Eot (2), and Eot (3) are September, respectively. , August and July are the average outside temperatures. Further, the model characteristic quantities SZ (1), SZ (2), and SZ (3) are saturated values of the functions Mp (1), Mp (2), and Mp (3) (parameters a (1) and a (2), respectively. ), A (3)).

Then, as shown in FIG. 14 (B), past outdoor thermal environment representative values {Eot (1), Eot (2), Eot (3)} and model characteristic quantities {SZ (1), SZ (2), Let the characteristic quantity correction function Gpm be a function that linearly approximates the relationship with SZ (3)}. The characteristic quantity correction function Gpm is a function for estimating the model characteristic quantity SZ (saturation value of the reporter rate PPVh of "hot") corresponding to the arbitrary monthly outside temperature.

In this embodiment, linear approximation is used as a method for obtaining the characteristic quantity correction function Gpm, but the characteristic quantity correction function Gpm considers the influence of the seasonal transition on the feeling of heat by a method such as polynomial approximation or exponential approximation. And it may be decided appropriately.

In this embodiment, the characteristic quantity correction function Gpm is determined by using the saturation value of the reporter rate PPVh of "hot" as the model characteristic quantity SZ, but the model characteristic quantity SZ is another that determines the shape of the function of the past model. The amount may be used, and a plurality of types of model characteristic quantity SZ may be set.

For example, the saturation value of the filer rate PPVh of "hot" is the model characteristic amount SZ1 (SZ in FIG. 13), and the PMV value of the rising position of the filer rate PPVh is the model characteristic amount SZ2 as shown in FIG. 15 (A). As shown in 15 (B), the PMV value when the filer rate PPVh reaches 50% of the saturation value is defined as the model characteristic quantity SZ3.

In the example of FIG. 15 (A), the model characteristic quantity SZ2 is set to e of the model search parameters a, b, c, d, and e when the individual thermal feeling models are the equations (15) and (16). Correspond.
Therefore, the season-dependent management unit 21 acquires the model characteristic quantity SZ2 (i) = e (i) from each of the modeling results Sp (i) of N_ref times, and similarly to the case of the model characteristic quantity SZ1, the model characteristic quantity. A function that linearly approximates the relationship between SZ2 (i) = e (i) and the past outdoor thermal environment representative value Eot (i) may be obtained, and this function may be set as the characteristic quantity correction function Gpm_2 (step S504).

Further, in the example of FIG. 15B, the model characteristic quantity SZ3 is a PMV value when the filer rate PPVh reaches 50% of the saturation value. Therefore, the season-dependent management unit 21 calculates the value a_50 (i) of 50% of a (i) from each of the model function of the individual thermal feeling model and the modeling result Sp (i) of N_ref times, and is "hot". The PMV value at which the reporter rate PPVh is a_50 (i) is calculated from the equation (16) (or the equation (15)) and the modeling result Sp (i), and the calculated result is referred to as SZ3 (i). Just do it.

Then, the seasonal dependence management unit 21 obtains a function that linearly approximates the relationship between the model characteristic quantity SZ3 (i) and the past outdoor thermal environment representative value Eot (i), as in the case of the model characteristic quantity SZ1. This function may be a characteristic quantity correction function Gpm_3 (step S504). Let Gpm_1 be the characteristic quantity correction function obtained from the model characteristic quantity SZ1 (SZ in FIG. 13).
In this way, the characteristic quantity correction function can be obtained for each type of model characteristic quantity SZ.

Finally, the seasonal dependence management unit 21 transmits the characteristic quantity correction function Gpm (Gpm_1, Gpm_2, Gpm_3) and the modeling result of the latest past model in response to the information transmission request from the future model generation unit 22 (Fig.). 11 steps S404).

The reference model number N_ref, the method for determining the past outdoor thermal environment representative value Eot, the method for determining the model characteristic quantity SZ, and the method for determining the characteristic quantity correction function Gpm corresponding to each characteristic quantity include indoor environment evaluation / control, etc. It is preset by the solution provider and equipment manager of.

FIG. 16 is a flowchart illustrating the operation of the future model generation unit 22. The future model generation unit 22 wants to estimate the feeling of heat of the residents of the evaluation target space 101 (the ratio of the residents who report dissatisfaction with the thermal environment of the evaluation target space 101, and the reporter rate PPVh of "hot"). (Estimated future date and time Tm) is acquired by an instruction from the operator or the like (FIG. 16 step S600).

The future model generation unit 22 acquires information on the update timing of the individual thermal realization model from the data period management unit 14, and determines the future learning data period LPF including the estimated future date and time Tm (step S601 in FIG. 16). For example, assuming that the estimated future date and time Tm is October 15, 2019, the update timing of the individual thermal realization model is 0:00 on the first day of every month, and the training data period width is one month. Therefore, the future model generation unit 22 Determines the future learning data period LPF as October 2019.

Then, the future model generation unit 22 acquires the future outdoor thermal environment information corresponding to the future learning data period LPF from the outdoor thermal environment information management unit 20 (step S602 in FIG. 16), and the acquired future outdoor thermal environment. Based on the information, the future outdoor thermal environment representative value Eotf corresponding to the future learning data period LPF is determined (FIG. 16 step S603). For example, when the future learning data period LPF is October 2019, the future model generation unit 22 may acquire the average outside temperature of the average month of October and set it as the future outdoor thermal environment representative value Etf.

Next, the future model generation unit 22 uses the characteristic quantity correction function Gpm (here, Gpm_1) transmitted from the seasonal dependence management unit 21 to model characteristic quantity SZ1 = corresponding to the future outdoor thermal environment representative value Eotf. Gpm_1 (Eoft) is calculated, and the calculated value is used as the estimated characteristic quantity SZest (step S604 in FIG. 16).

Finally, the future model generation unit 22 generates an individual thermal realization model (future model) in which the latest past model is modified based on the calculated estimated characteristic quantity SZest (step S605 in FIG. 16). The method of modifying the past model is preset by the solution provider such as indoor environment evaluation / control and the equipment manager.

In this embodiment, the ratio between the estimated characteristic quantity SZest and the model characteristic quantity SZ1 (1) = a (1) of the latest past model is set as the model correction parameter η.
η = SZest / SZ1 (1) ... (17)

Then, the future model generation unit 22 modifies the function Mp (1) of the latest past model based on the model correction parameter η as follows.

In the equation (18), among the parameters a (1), c (1), d (1), and e (1) of the latest past model (equation (16)), the parameter a (1) is used as the model correction parameter η. It is shown to be corrected based on.

In the modification of the past model of this embodiment, only the magnification of the PPVh axis is modified by the correction parameter η, but by using the above model characteristic quantities SZ2 and SZ3, the rising position of the filer rate PPVh The correction or the magnification correction of the PMV axis may be performed.

Specifically, the future model generation unit 22 uses the characteristic quantity correction function Gpm_2 transmitted from the seasonal dependence management unit 21 to model characteristic quantity SZ2 = Gpm_2 (Eoft) corresponding to the future outdoor thermal environment representative value Eotf. ) Is calculated, and the calculated value is used as the estimated characteristic quantity SZest_2. Further, the future model generation unit 22 calculates the model characteristic quantity SZ3 = Gpm_3 (Eoft) corresponding to the future outdoor thermal environment representative value Eotf by using the characteristic quantity correction function Gpm_3, and estimates the calculated value. It is set to SZest_3 (step S604 in FIG. 16).

Then, for example, the difference between the estimated characteristic quantity SZest_2 and the model characteristic quantity SZ2 (1) = e (1) of the latest past model is set as the model correction parameter δ.
δ = SZest_2-SZ2 (1) ... (19)

Further, for example, the difference between the estimated characteristic quantities SZest_3 and SZest_2 (SZest_3-SZest_2) and the difference between the model characteristic quantities SZ3 (1) and SZ2 (1) of the latest past model (SZ3 (1) -SZ2 (1)). Let the ratio with and be the model correction parameter ξ.
ξ = (SZest_3-SZest_2) / (SZ3 (1) -SZ2 (1))
... (20)

Then, the future model generation unit 22 modifies the function Mp (1) of the latest past model based on the model correction parameters η, δ, ξ as follows.

By modifying the past model as in equation (21), it is possible to modify the rising position of the filer rate PPVh by the model correction parameter δ and to modify the magnification of the PMV axis by the model correction parameter ξ.

As a result, the processing of the future model generation unit 22 is completed. According to this embodiment, by modifying the past model based on the outdoor thermal environment condition reflecting the seasonal transition to generate the future model, the estimation error due to the seasonal transition element can be improved, and the future model can be used. If the estimated PMV value of the estimated future date and time is input, the resident's feeling of warmth with respect to the evaluation target space 101 at the estimated future date and time (the ratio of the resident who declares dissatisfaction with the thermal environment of the evaluation target space 101, which is called "hot". It is possible to estimate the filer rate PPVh). As a result, in this embodiment, the estimation result of the resident's feeling of heat can be brought closer to the actual situation.

[Second Example]
Next, a second embodiment of the present invention will be described. Also in this embodiment, the configuration of the thermal sensation model generator is the same as that in the first embodiment, and thus the reference numerals in FIG. 2 will be used for description.
If training data corresponding to a period of several days including short-term fluctuations is used, the reliability of the model cannot be maintained, so it is necessary to secure data for a certain period as training data (integrated information). As described in the first embodiment, the learning data period width is appropriately about one month, but in order to maintain the continuity of the seasonal transition, an overlapping period in which the individual learning data periods overlap is provided. In this embodiment, an example in which the overlap period is set to 2 weeks as a reliable constant period will be described.

The data period management unit 14 of the thermal sensation model generator of this embodiment has the data period management for determining the update timing and the learning data period of the individual thermal sensation model (past model) as in the first embodiment. Information is preset by solution providers such as indoor environment evaluation / control and equipment managers.

The difference from the first embodiment is that the data period management information includes the information of the overlap period width Lo in addition to the information of the update timing of the individual thermal feeling model and the information of the learning data period width.

17 (A) and 17 (B) are diagrams for explaining the learning data period of the first and second embodiments. In addition, in FIGS. 17 (A) and 17 (B), the update timing dates are uniformly set to 30 days for each month for easy understanding, and October 1, 2018 is the latest update timing. The learning data period of each update timing that goes back to the update timing of August 1, 2018 is shown. Further, in both FIGS. 17 (A) and 17 (B), the update start date and time is set to 0:00 on August 1, 2018, and the period width of the learning data is set to every 30 days. In the example of FIG. 17 (B), The overlap period width Lo of this embodiment is set to 15 days. In FIG. 17 (B) of the present embodiment in which the overlap period width Lo is set for the learning data period of continuous update timing as shown in FIG. 17 (A), the model update cycle is set to the overlap period width Lo. Correspondingly shortened.

In the example of this description corresponding to the first embodiment, as shown in FIG. 17 (A), for example, the training data period SW of the past model Mp (2) updated at 0:00 on September 1, 2018. Is 30 days with the start timing of 0:00 on August 1st and the end timing of 23:59 on August 30th (here, since each month is uniformly set to 30th, it corresponds to the end of August). is there. Furthermore, the learning data period SW of the previous past model Mp (3) starts at 0:00 on July 1st and is July 30th (here, each month is uniformly 30th, so the end of July). It is 30 days with 23:59 as the end timing (corresponding to the day), and the learning data period SWs of the past models Mp (2) and Mp (3) of the continuous update timing do not overlap.

On the other hand, in this embodiment, as shown in FIG. 17B, for example, the training data period SW of the past model Mp (3) updated at 0:00 on September 1, 2018 is 0 on August 1. The start timing is 0:00, and the end timing is August 30 (corresponding to the end of August because each month is uniformly set to 30 days), which is the 30th with the end timing shown in FIG. 17 (A). The training data period SW of the past model Mp (2) is the same as that of the past model Mp (2), but the training data period SW of the previous model Mp (4) in FIG. 17 (B) starts at 0:00 on July 16. The timing includes an overlap period width Lo that overlaps with the training data period SW of the past model Mp (3) of FIG. 17 (B) (that is, 15 days from August 1 to August 15), and August 15 It is 30 days with the end timing at 23:59 on the day, and the learning data period SWs of the past models Mp (3) and Mp (4) of the continuous update timing overlap by the overlap period width Lo.

Other configurations are as described in the first embodiment. It should be noted that the learning data period width of this embodiment is appropriately about one month, but at this time, in order to maintain the continuity of the seasonal transition, the reliable period width overlaps with the learning data period before and after. It is desirable to let it. It is appropriate that the overlap period width Lo is a long period of 2 weeks or more.

The two-week average transition of the outside air temperature measured from July to September 2017 in a certain building is shown in FIG. 18 (A), and the four-week average transition is shown in FIG. 18 (B). The horizontal axis of FIGS. 18A and 18B is the number of elapsed weeks counted from the first week.

In the example of FIG. 18 (A), the overlap period is one week, and in the example of FIG. 18 (B), the overlap period is two weeks. As is clear from the comparison of FIGS. 18 (A) and 18 (B), when the overlap period is 2 weeks and the average value of 4 weeks is taken, it is difficult for the average value of the outside air temperature to include short-term fluctuations. I understand.

The thermal sensation model generation device described in the first and second embodiments can be realized by a computer provided with a CPU (Central Processing Unit), a storage device, and an interface, and a program for controlling these hardware resources. .. A configuration example of this computer is shown in FIG.

The computer includes a CPU 200, a storage device 201, and an interface device (hereinafter, abbreviated as I / F) 202. An environment measurement device or the like is connected to the I / F 202. In such a computer, a program for realizing the method for generating a thermal sensation model of the present invention is stored in the storage device 201. The CPU 200 executes the processes described in the first and second embodiments according to the program stored in the storage device 201.

The present invention can be applied to a technique for estimating a resident's feeling of heat at a future date and time.

1 ... Past model generation unit, 2 ... Future model generation unit, 10 ... Data collection unit, 11 ... Living environment information holding unit, 12 ... Thermal feeling information holding unit, 13 ... Data integration unit, 14 ... Data period management department, 15 ... Past model generation unit, 16 ... Past model memory management department, 20 ... Outdoor thermal environment information management department, 21 ... Seasonal dependence management department, 22 ... Future model generation department.

Claims (10)

Data configured to store the update timing information and the learning data period information of the first model that models the relationship between the living environment index of the evaluation target space and the resident's feeling of heat with respect to the evaluation target space. With the period management department
When the update timing comes, the first model generation unit configured to generate the first model based on the living environment index in the learning data period and the thermal feeling information of the resident. ,
The function of the first model is based on the plurality of first models generated by the plurality of training data periods and the representative values of the past outdoor thermal environments corresponding to each of the plurality of training data periods. A seasonally dependent management unit configured to obtain a characteristic quantity correction function that indicates the relationship between the model characteristic quantity that determines the shape and the representative value of the outdoor thermal environment.
The future learning data period to be set for the future date and time for which the resident's feeling of heat is to be estimated is determined, and the model characteristic amount corresponding to the representative value of the outdoor thermal environment in this learning data period is defined as the characteristic amount. It is provided with a second model generation unit configured to generate a second model obtained by calculating based on a modification function and modifying the latest first model based on the calculated model characteristic quantity. A featured thermal sensation model generator. In the thermal feeling model generator according to claim 1,
The data period management unit stores information on the overlap period in addition to the information on the update timing and the information on the learning data period.
The plurality of learning data periods are set so that the learning data periods before and after the continuation overlap by the width of the overlapping period. In the thermal feeling model generator according to claim 1 or 2.
It is further equipped with an outdoor thermal environment information management unit configured to store past outdoor thermal environment information of the building including the evaluation target space.
The seasonally dependent management unit acquires the past outdoor thermal environment information corresponding to each of the plurality of learning data periods from the outdoor thermal environment information management unit, and the past corresponding to each of the plurality of learning data periods. The representative value of the outdoor thermal environment was determined, and the characteristic quantity correction function was obtained.
The second model generation unit acquires the past outdoor thermal environment information corresponding to the future learning data period from the outdoor thermal environment information management unit, and obtains the outdoor thermal environment corresponding to the future learning data period. A thermal sensation model generator characterized by determining a representative value of. In the thermal feeling model generator according to any one of claims 1 to 3.
The second model generation unit modifies the latest first model based on the model characteristic quantity calculated based on the characteristic quantity correction function and the latest model characteristic quantity of the first model. A thermal sensation model generator featuring. In the thermal feeling model generator according to any one of claims 1 to 4.
The living environment index is PMV, and the resident's feeling of heat is the proportion of the resident who declares dissatisfaction with the heating environment. Refer to the data period management unit that stores the update timing information and the learning data period information of the first model that models the relationship between the living environment index of the evaluation target space and the resident's feeling of heat with respect to the evaluation target space. Then, when the update timing is reached, the first step of generating the first model based on the living environment index in the learning data period and the thermal feeling information of the resident,
The function of the first model is based on the plurality of first models generated by the plurality of training data periods and the representative values of the past outdoor thermal environments corresponding to each of the plurality of training data periods. The second step of finding the characteristic quantity correction function showing the relationship between the model characteristic quantity that determines the shape and the representative value of the outdoor thermal environment, and
The future learning data period to be set for the future date and time for which the resident's feeling of heat is to be estimated is determined, and the model characteristic amount corresponding to the representative value of the outdoor thermal environment in this learning data period is defined as the characteristic amount. Thermal sensation model generation, which includes a third step of calculating based on a modification function and generating a second model modified from the latest first model based on the calculated model characteristic quantity. Method. In the method for generating a thermal feeling model according to claim 6,
The data period management unit stores information on the overlap period in addition to the information on the update timing and the information on the learning data period.
The method for generating a thermal sensation model, wherein the plurality of learning data periods are set so that the learning data periods before and after the continuation overlap by the width of the overlapping period. In the method for generating a thermal sensation model according to claim 6 or 7.
The second step refers to the outdoor thermal environment information management unit that stores the past outdoor thermal environment information of the building including the evaluation target space, and refers to the past outdoor thermal environment corresponding to each of the plurality of learning data periods. Including the step of acquiring the thermal environment information from the outdoor thermal environment information management unit, determining the representative value of the past outdoor thermal environment corresponding to each of the plurality of learning data periods, and obtaining the characteristic quantity correction function.
In the third step, the past outdoor thermal environment information corresponding to the future learning data period is acquired from the outdoor thermal environment information management unit, and the representative of the outdoor thermal environment corresponding to the future learning data period is obtained. A method for generating a thermal sensation model, which comprises a step of determining a value. In the method for generating a thermal sensation model according to any one of claims 6 to 8.
The third step includes a step of modifying the latest first model based on the model characteristic quantity calculated based on the characteristic quantity correction function and the latest model characteristic quantity of the first model. A method for generating a thermal sensation model, which is characterized in that. In the method for generating a thermal sensation model according to any one of claims 6 to 9.
The method for generating a thermal sensation model, wherein the living environment index is PMV, and the resident's thermal sensation is the proportion of resident who declares dissatisfaction with the thermal environment.