JP6538487B2 - Heterogeneous energy consumption estimation device and heterogeneous characteristic estimation device - Google Patents

Description

  The present invention relates to a heterogeneous energy consumption estimation device that estimates consumption of a certain energy of a consumer who consumes multiple types of energy such as city gas and electricity from consumption of another energy, and in particular, a city gas company And estimate the amount of consumption of one energy of a group of general household consumers who receive and consume energy supply such as city gas for home use or electricity from a power company etc. from the consumption of another energy. The present invention relates to an energy consumption estimation device.

  Conventionally, various things have been proposed regarding estimation of energy consumption of power, city gas, etc. or energy demand prediction (see, for example, Patent Documents 1 to 4 listed below). Patent Document 1 proposes a method of estimating the consumption per unit time within an arbitrary period for each customer. Patent Document 2 proposes an energy demand prediction system in the case where energy is supplied to a plurality of energy consumers from an energy production site through an energy supply network. Patent Document 3 proposes a power consumption management device that estimates the power consumption in model units of power consumption devices. Patent Document 4 proposes a power consumption estimation system for high load devices in the nighttime time zone. In any of Patent Documents 1 to 4, estimation and the like of energy consumption are performed based on actual data of the same type of energy, that is, estimation of power consumption is performed based on the history data of power consumption. There is.

Unexamined-Japanese-Patent No. 2006-11715 JP 2007-241360 A JP 2012-90502 A JP 2014-138455 A

  As described above, in the conventional estimation of energy consumption or energy demand prediction, the energy species to be estimated or predicted and the energy species of the actual data used for the estimation or prediction are the same. Conventionally, it has been considered difficult to estimate different types of energy consumption (for example, city gas consumption from power consumption) from certain energy consumption. It is a factor that the energy consumption pattern in the general household is largely different depending on the energy type.

  Also, conventionally, there has been almost no demand for estimating disparate energy consumption from a certain energy consumption. However, with the liberalization of the energy market, such as electricity and city gas, a supplier of only one energy type, such as a power company or a city gas company, can enter into the supply of different types of energy, and further, the supply of different types of energy. However, it is important to estimate the consumption of such disparate energy, as it extends to a wide range of household consumers.

  An object of the present invention is to provide a heterogeneous energy consumption estimation apparatus capable of estimating heterogeneous energy consumption from a certain energy consumption with a practical accuracy in which estimation errors are suppressed.

In order to achieve the above object, the heterogeneous energy consumption estimation apparatus according to the present invention is characterized in that the first period per consumer in the group of consumers consuming different first energy and second energy is different Based on known sample data in which a pair of consumption of one energy and consumption of the second energy in the same or different second period as the first period is used as individual sample values, the known period of the first period is obtained. It is a heterogeneous energy consumption estimation device for estimating the consumption of the unknown second energy of the second period from the consumption of the first energy,
Storage means for storing the sample data, and one or more first threshold values as boundary values for dividing the distribution range of consumption of the first energy in the first period into a plurality of first segments, the second period First threshold calculation means for calculating the distribution range of consumption of the second energy corresponding to one or more preset second thresholds which are boundary values for dividing the distribution range of the second energy into a plurality of second segments, and , And
The first threshold value calculation means, for all of the first segments or for each of preselected ones, corresponds to the first one of the total number of the sample values belonging to one first segment. A first feature is that each of the first threshold values is calculated such that the hit rate, which is a ratio of the number of sample values belonging to two segments, is equal to or greater than a predetermined value.

  According to the first feature configuration, in all of the first segments or each of the portions selected in advance, the sample values distributed in the first segments correspond to the corresponding values with an accuracy rate equal to or higher than the predetermined value. Since it exists in two segments, when estimating the consumption of the 2nd energy of the consumer of the consumption of the 1st energy of the 1st segment concerned, the consumption of the 2nd energy of the corresponding 2nd segment Therefore, the estimation accuracy of the consumption of the second energy can be suppressed within a certain error range. That is, the estimation error can be reduced as compared with the case where the distribution ranges of the consumption amounts of the first and second energy are not divided into the first and second segments, respectively.

  In the heterogeneous energy consumption amount estimation apparatus having the first characteristic configuration, the first threshold value calculation means determines that the accuracy rate is equal to or more than the predetermined value with respect to each of a predetermined portion of the first segment. Calculating each of the first threshold values so that the number of target samples, which is the number of the sample values belonging to both the one first segment and the one second segment corresponding thereto, is maximized within The second feature.

  According to the second feature configuration, it is possible to uniquely define the first threshold that defines the range of the first segment for which the hit ratio is equal to or more than the predetermined value. In addition, the estimation error of the consumption amount of the second energy to the first segment can be further suppressed by maximizing the number of hits.

  In the heterogeneous energy consumption amount estimation apparatus having the first or second characteristic configuration, the aggregate is divided into a plurality of groups based on one or more attributes of the consumer, and the storage means is configured to be the individual. To store the sample data so that it can be identified to which of the plurality of groups the sample value belongs, and one or more of the second threshold values are commonly set through the plurality of groups, and the first threshold value is stored. A third feature is that the calculating means calculates the first threshold corresponding to the second threshold separately for each of the groups.

  According to the third feature configuration, since the definition of the first segment is made more appropriately according to the attribute of the consumer for each group, the estimation error of the consumption of the second energy can be further suppressed.

  The different energy consumption estimation apparatus having the first or second characteristic configuration receives, as an input, the consumption of the first energy of the first period of the one consumer selected from the group, and The second energy consumption calculating unit is configured to calculate the consumption of the second energy in the second period by using a calculation formula set for every two segments, and the second energy consumption calculating unit has an input as an input. The amount of consumption of the received first energy is compared with the first threshold value to determine which of the plurality of first segments it belongs to, and the second corresponding to the determined first segment A fourth feature is to calculate the consumption amount of the second energy in the second period using the calculation formula set for the segment.

In the heterogeneous energy consumption estimation apparatus having the third characteristic configuration, the consumption amount of the first energy of the first period of the consumer selected from the aggregate and the group to which the consumer belongs A second energy consumption calculation unit that receives identification information as an input and calculates the consumption of the second energy in the second period using a calculation formula set for each of the second segments,
The second energy consumption amount calculation means compares the consumption amount of the first energy received as an input with the first threshold value calculated for the group of the identification information received as an input. It is determined which of the first segments it belongs to, and using the calculation formula set for the second segment corresponding to the determined first segment, the second of the second period The fifth feature is to calculate energy consumption.

  According to the fourth or fifth aspect, when estimating the consumption of the second energy corresponding to the consumption of the first energy, the first segment to which the consumption of the first energy belongs belongs to the first segment. In order to estimate the consumption of the second energy by using a calculation formula specialized to the second segment, the corresponding second segment is determined with a hit rate higher than a predetermined value, and therefore, the estimation of the consumption of the second energy Accuracy can be kept within a certain error range. That is, the effects of the first to third feature configurations can be more specifically exhibited.

Furthermore, in the heterogeneous energy consumption estimation apparatus according to the present invention, it is preferable that the first energy of the first period for each consumer is known in the group of consumers consuming different first energy and second energy. What is claimed is: 1. A different energy consumption estimation device for estimating the consumption of the unknown second energy in a second period that is the same as or different from the first period from the consumption.
Storage means for storing one or more first threshold values which are boundary values for dividing the distribution range of consumption of the first energy in the first period into a plurality of first segments, and one selected from the group A second energy consumption calculation for receiving the consumption of the first energy of the first period of the consumer as an input and calculating the consumption of the second energy for the second period using a predetermined calculation formula Equipped with
The pair of the known consumption of the first energy in the first period and the consumption of the second energy in the second period of each consumer in the assembly is taken as an individual sample value;
The first threshold corresponds to one or more second thresholds, which are boundary values dividing the distribution range of the consumption of the second energy of the second period into a plurality of second segments, Of the number of sample values belonging to the corresponding one second segment among the total number of sample values belonging to one of the first segments, for each of all or a portion selected in advance. It is set so that a certain hit rate is equal to or more than a predetermined value,
The calculation formula is set for each of the second segments,
The second energy consumption calculation means compares the consumption of the first energy received as an input with the first threshold to determine which of the plurality of first segments the energy consumption belongs to, A sixth feature is to calculate the consumption amount of the second energy in the second period, using the calculation formula set for the second segment corresponding to the determined first segment.

Furthermore, in the heterogeneous energy consumption estimation apparatus according to the present invention, it is preferable that the first energy of the first period for each consumer is known in the group of consumers consuming different first energy and second energy. What is claimed is: 1. A different energy consumption estimation device for estimating the consumption of the unknown second energy in a second period that is the same as or different from the first period from the consumption.
The assembly is divided into a plurality of groups based on one or more attributes of the consumer,
One or more first thresholds, which are boundary values for dividing the distribution range of the consumption of the first energy in the first period into a plurality of first segments, and each of the consumers belongs to any of the plurality of groups Storage means for storing distinguishable information, the consumption amount of the first energy in the first period of the one consumer selected from the group, and identification information of the group to which the consumer belongs And second energy consumption calculation means for calculating the consumption of the second energy in the second period using a predetermined calculation formula.
The pair of the known consumption of the first energy in the first period and the consumption of the second energy in the second period of each consumer in the assembly is taken as an individual sample value;
The first threshold corresponds to one or more second thresholds, which are boundary values dividing the distribution range of the consumption of the second energy of the second period into a plurality of second segments, Of the number of sample values belonging to the corresponding one second segment among the total number of sample values belonging to one of the first segments, for each of all or a portion selected in advance. It is set according to the above-mentioned group so that a certain hit rate will be more than a predetermined value,
The calculation formula is set for each of the second segments,
The second energy consumption amount calculation means compares the consumption amount of the first energy received as an input with the first threshold set for the group of the identification information received as an input. It is determined which of the first segments it belongs to, and using the calculation formula set for the second segment corresponding to the determined first segment, the second of the second period A seventh feature is to calculate energy consumption.

  The different energy consumption amount estimation apparatus having the sixth or seventh characteristic configuration is characterized in that the first threshold is higher than the predetermined value with respect to each of preselected portions of the first segment. In the range where the number of sample values belonging to both the one first segment and the one corresponding second segment is the maximum number, the eighth being that it is set It features.

  According to the sixth to eighth feature configurations, as in the fourth or fifth feature configurations, when estimating the consumption amount of the second energy corresponding to the consumption amount of the first energy, The second segment corresponding to the first segment to which the consumption amount of the first energy belongs by the hit rate above the predetermined value is determined, and the consumption amount of the second energy using the calculation formula specialized for the second segment The estimation accuracy of the consumption amount of the second energy can be suppressed within a certain error range.

  In the heterogeneous energy consumption estimation apparatus having any one of the fourth to eighth characteristic configurations, the calculation formula uses the consumption of the second energy in the second period as a target variable, and A ninth feature is that the regression equation is set for each of the second segments, the amount of consumption of 1 energy being an explanatory variable.

  In the heterogeneous energy consumption estimation apparatus of the ninth feature configuration, the regression equation further includes at least one of one or more attributes of one of the consumers selected from the aggregate as an explanatory variable. A tenth feature is that it is a regression equation.

  In the heterogeneous energy consumption estimation device having the ninth or tenth feature configuration, the regression equation is a regression equation derived by regression analysis using the sample value belonging to the corresponding second segment. 11 features.

  The heterogeneous energy consumption estimation apparatus having any one of the fourth to eleventh characteristic configurations selects the consumers one by one from the aggregate, and selects the first energy of the first period of the consumers. A second energy consumption tabulation summarizing consumption of the second energy in the second period, which is input to the second energy consumption calculating means and sequentially output from the second energy consumption calculating means. A twelfth feature is to provide means.

Furthermore, in the heterogeneous characteristic estimating apparatus according to the present invention, the first characteristic value of one of the two events and the second characteristic value of the other of two sets of individuals exhibiting predetermined characteristics with respect to two different events respectively. What is claimed is: 1. A heterogeneous characteristic estimation apparatus for estimating an unknown second characteristic value from a known first characteristic value based on known sample data in which a pair is an individual sample value,
Storage means for storing the sample data, at least one first threshold which is a boundary value dividing the distribution range of the first characteristic value into a plurality of first segments, and a plurality of distribution ranges of the second characteristic value First threshold calculation means for respectively calculating one or more preset second threshold values which are boundary values to be divided into second segments,
The first threshold value calculation means, for all of the first segments or for each of preselected ones, corresponds to the first one of the total number of the sample values belonging to one first segment. A first feature is that each of the first threshold values is calculated such that the hit rate, which is a ratio of the number of sample values belonging to two segments, is equal to or greater than a predetermined value.

  According to the first feature configuration, in all of the first segments or each of the portions selected in advance, the sample values distributed in the first segments correspond to the corresponding values with an accuracy rate equal to or higher than the predetermined value. Since it exists in 2 segments, when estimating the 2nd characteristic value of the individual of the 1st characteristic value of the 1st segment concerned, probability of existing in the range of the 2nd characteristic value of the corresponding 2nd segment Of the second characteristic value can be suppressed within a certain error range. That is, the estimation error can be reduced as compared with the case where the distribution ranges of the first and second characteristic values are not divided into the first and second segments, respectively.

  In the heterogeneous characteristic estimating apparatus of the first characteristic configuration, the first threshold value calculation means may set the accuracy prediction range to be equal to or more than the predetermined value with respect to each of the previously selected ones of the first segments. Second calculation of the first threshold so that the number of target samples, which is the number of the sample values belonging to both the one first segment and the one second segment corresponding thereto, is maximized, It is the feature of

  According to the second feature configuration, it is possible to uniquely define the first threshold that defines the range of the first segment for which the hit ratio is equal to or more than the predetermined value. In addition, the estimation error of the second characteristic value with respect to the first segment can be further suppressed by maximizing the number of hits.

  In the heterogeneous characteristic estimation apparatus having the second or third feature configuration, the aggregate is divided into a plurality of groups based on one or more attributes of the individual, and the storage unit is configured to calculate the individual sample values. Stores the sample data so that it can be identified to which of the plurality of groups, one or more of the second threshold is commonly set through the plurality of groups, and the first threshold calculation unit A third feature is to calculate the first threshold corresponding to the second threshold separately for each group.

  According to the third feature, since the definition of the first segment is made more appropriate for each group according to the attribute of the individual, the estimation error of the second characteristic value can be further suppressed.

  According to the heterogeneous energy consumption estimation apparatus having the above-described feature configuration, it is possible to estimate heterogeneous energy consumption from a certain energy consumption with a practical accuracy in which estimation errors are suppressed.

A block diagram showing a schematic configuration of an embodiment of the heterogeneous energy consumption estimation apparatus according to the present invention A scatter plot of the monthly values of city gas consumption and power consumption of individual consumers plotted as sample points A flowchart showing an outline processing procedure of the first processing by the first threshold value calculation means A diagram schematically illustrating the processing content of step # 12 of the first processing shown in FIG. 3 A diagram schematically showing an example of the processing result of the first processing A flowchart showing a schematic processing procedure of the second processing by the power consumption calculation means A table showing the results of evaluating the estimation error by the device of the present invention according to the presence or absence of the first processing, the presence or absence of the group classification, and the difference in calculation formula A table showing the error improvement effect due to the first process, the group classification, and the calculation formula for the estimation error shown in FIG. 7

  Hereinafter, an embodiment of a heterogeneous energy consumption estimation apparatus (hereinafter, appropriately referred to as a “present invention apparatus”) according to the present invention will be described with reference to the drawings.

  The device according to the present invention is, for a population (population) of consumers consuming a plurality of types of energy such as city gas, electric power, etc., the first energy of the plurality of types of energy in a first period (for example, And the consumption amount of the second energy (for example, power) of the plurality of types of energy in the same or another second period as the first period. In the following description, city gas as the first energy and power as the second energy are assumed, and one month is assumed as the first and second periods, and the first and second periods are the same month Assume. That is, for example, when the first period is X months (X = 1 to 12), the second period is also X months. In addition, as consumers consuming multiple types of energy, it is assumed that general household households are supplied with city gas and electric power for general households.

  FIG. 1 is a block diagram showing an example of the configuration of the device 1 of the present invention. As shown in FIG. 1, the device 1 of the present invention includes a storage unit 2, a first threshold calculation unit 3, a power consumption calculation unit 4, a power consumption totaling unit 5, and a calculation formula derivation unit 6. ing. The device 1 of the present invention is configured using, for example, hardware resources (CPU, storage device, interface device, etc.) and software resources (OS, program, etc.) of a general-purpose computer such as a personal computer, and the first threshold value calculation means 3. The power consumption calculation means 4, the power consumption totaling means 5, and the calculation formula derivation means 6 are provided in the computer, and the software is executed by executing a program for processing each part separately. To be realized. The storage unit 2 may be configured using a non-volatile storage device (hard disk drive (HDD), solid state drive (SSD), etc.) built in a general-purpose computer, but sharing stored data with other systems In this case, it may be provided as a database outside the general-purpose computer and connected via a data communication channel such as a local area network (LAN). The power consumption calculating means 4 and the power consumption collecting means 5 correspond to a second energy consumption calculating means and a second energy consumption collecting means.

  In the present embodiment, the data stored in the storage device 1 includes the consumer identification code, the actual value or estimated value of the city gas consumption for each month, and the power consumption for each month for each individual consumer of the above aggregate. The actual value or estimated value, and the group number which is identification information of the group indicating the attribute classification of the consumer. Here, a pair of actual values of city gas consumption and power consumption of each month for each consumer is referred to as a sample value, and a set of the sample values of the entire assembly is referred to as sample data.

  In the present embodiment, each consumer in the collection is classified into a plurality of groups according to some attributes of the consumer. As the attribute, for example, a residence type (household or apartment house), a family configuration, the presence or absence of a specific power or gas consumption device, the presence or absence of a distributed power generation device (such as a solar panel), etc. are assumed. The number of attributes used and the total number of groups are arbitrary, but if the number of groups is too large relative to the total number of consumers, the number of samples in one group will be small, so the number of samples in each group will be too small Adjust so as not to. Incidentally, since the appropriate range of the number of samples in each group is also related to the number of segments described later, it is preferable to set appropriately in accordance with the characteristics of the assembly to be used.

  In FIG. 2, the city gas which is a sample value of an individual consumer in a two-dimensional coordinate space where the vertical axis is the normalized power consumption of a month and the horizontal axis is the normalized city gas consumption of the same month. The scatter diagram which plotted the value which normalized consumption and power consumption as a sampling point is shown. In the example shown in FIG. 2, the number of sampling points is about 8000. Normalized power consumption and normalized city gas consumption are values obtained by dividing the power consumption and city gas consumption of each sample value by the maximum values of the power consumption and city gas consumption of each sample value. It is. Hereinafter, the normalized power consumption and the normalized city gas consumption will be appropriately referred to as the power consumption E and the city gas consumption G, respectively. In the example shown in FIG. 2, the power consumption E and the city gas consumption G are in the range where the power consumption E is approximately 0.02 to 0.4 and the city gas consumption G is approximately 0 to 0.45. It can be seen that it clearly shows positive correlation.

  Next, the content of the process (first process) of the first threshold value calculation unit 3 will be described with reference to FIGS. 3 to 5. FIG. 3 is a flowchart showing the processing procedure of the first processing. FIG. 4 is a diagram for schematically explaining the processing contents of step # 12 shown in FIG.

  In the present embodiment, the power consumption to be estimated is divided into a plurality of segments (second segments) before the first process is performed. The boundary value of each second segment (second threshold value TEi, i = 1 to m-1, m is the number of segments so that sample points showing the above-mentioned positive correlation are included in each second segment after division ) Is set in advance. Note that the second threshold value TE0 = 0 and the second threshold value TEm = 1, and i is expanded to 0 to m. In the following description, assuming that power consumption is divided into four second segments for convenience, each second segment has power consumption E of 0 to TE1, TE1 to TE2, TE2 to TE3, TE3. It becomes four ways of ~. In the example shown in FIG. 2, in order to include the sample point showing the above-mentioned positive correlation in the second segment (TE3 to 1), the second threshold value TE3 should be set to around 0.2. become. TE1 and TE2 are also set on the same basis. In the present embodiment, the second threshold values TE1 to TE3 are common to the groups.

  The first process by the first threshold value calculation means 3 is a process of dividing the city gas consumption into the same number of segments (first segments) as the power consumption. When the segment number m is 4, each first segment has four city gas consumption amounts of 0 to TG1j, TG1j to TG2j, TG2j to TG3j, and TG3j to 1, for each group. j is an ordinal number indicating 1 to n (n is the total number of groups) indicating the distinction of groups. The first processing corresponds to first and second threshold values TG1j to TG (m-1) j, which are boundary values of the first segments, corresponding to second threshold values TE1 to TE (m-1), respectively. This is a process of determining each group so as to satisfy the second classification standard.

  The first process will be described below along the flow shown in FIG. In the following description, the numbers in the order in which the first segment and the second segment are arranged are 1, 2,... M in the order in which the city gas consumption G and the power consumption E increase from 0 to 1.

  First, at step # 10, variables i and j are initialized to (m-1) and 1, respectively. Subsequently, in step # 11, the first threshold value TGij is tentatively set to the initial value TG0ij. TG0ij is, for example, 1 in the case of i = m−1, and the first threshold TG (i + 1) j determined in the loop immediately before the variable i in the case of i = 1 to m−2. .

  Subsequently, at step # 12, the hit ratio Rij shown in the following equation 1 is calculated while decreasing the first threshold value TGij stepwise in predetermined small steps.

  (Equation 1) Rij = Nij / N0ij × 100 (%)

  N0 ij in the equation 1 is within the range of TG ij to TG 0 ij where the city gas consumption G is within the range (corresponding to the area obtained by adding the area A and the area B in FIG. 4), that is, the provisional (i + 1) th first segment In the total number of samples in, Nij is the power consumption E of which is within the range of TEi to TE (i + 1) (corresponding to the region A in FIG. 4 within the (i + 1) th second segment). It is a certain number of samples. Hereinafter, Nij will be referred to as the "target number". TE (i + 1) is 1 in the case of i = m-1.

  While the target predictive value Rij is equal to or higher than a predetermined reference value (for example, 70%) and is maintained at or above the reference value, N0ij and Nij are counted while stepwise decreasing the first threshold TGij. Calculate the rate Rij. Normally, since city gas consumption G and power consumption E show a positive correlation, if the first threshold TGij is decreased to widen the range of TGij to TG0ij, N0ij and Nij will increase, respectively. Since the degree of increase is large, the hit rate Rij decreases. Then, TGij which satisfies the first classification criterion that the accuracy ratio Rij is equal to or more than the reference value and satisfies the second classification criterion that the number Nij of hitting targets is the maximum is determined as the first threshold TGij. In addition, since various methods can be considered as an algorithm for finding out TG ij satisfying the first and second classification criteria, and since the algorithm is not the gist of the present invention, detailed description will be omitted.

  Note that the purpose of applying the second classification criterion is to estimate the error of the power consumption E by maximizing the number Nij of hits if the first threshold TGij satisfying the first classification criterion is not uniquely determined. And the first threshold value TGij is uniquely determined. However, if the first threshold TG ij satisfying the first classification criterion is not determined in the (i + 1) -th first segment, the second classification criterion is ignored in the immediately preceding (i + 2) -th first segment. In some cases, the first threshold TG (i + 1) j that has already been determined may be raised stepwise to be adjusted such that a first threshold TGij that satisfies the first classification criterion is determined in the (i + 1) th first segment. obtain.

  Subsequently, if i ≠ 1 is determined in step # 13, i is decremented by 1 in step # 14, and steps # 11 to # 14 are repeated until the variable i is determined to be 1, and the first group j is selected. All threshold values TGij (i = 1 to m-1) are derived.

  Subsequently, when it is determined in step # 13 that i = 1, it is determined whether or not the variable j is n in step # 15. If the variable j is not n, the variable j is incremented by 1 in step # 16. i is initialized to (m-1), and steps # 11 to # 14 are repeated to derive the first threshold TGij (i = 1 to m-1) of the next group j. Subsequently, steps # 11 to # 16 are repeated until the variable j is determined to be n in step # 15, and the first threshold TGij (i = 1 to 1) is applied to all the groups j (j = 1 to n). Derivate m-1).

  FIG. 5 schematically shows the processing result in one group when the number of segments m is four, as an example. By executing the first processing of the first threshold value calculation means 3, the relationship between the first segment of the city gas consumption and the second segment of the power consumption is determined for each group. Knowing the city gas consumption, it is possible to identify to which second segment the power consumption of the consumer falls with a certain probability. Further, by increasing the number of segments m, the resolution of the first segment of i = 2 to m-2 is increased, so that the accuracy of the estimation of the power consumption is improved.

  Next, the content of the process (second process) of the power consumption calculation unit 4 will be described with reference to FIG. FIG. 6 is a flowchart showing the processing procedure of the second processing.

  At step # 20, the city gas consumption of one consumer is accepted together with the group number, and first, the group to which the consumer belongs is specified. Furthermore, the city gas consumption amount G is calculated by normalizing the received city gas consumption amount. When the city gas consumption amount G is received instead of the city gas consumption amount, the normalization is unnecessary. In addition, the city gas consumption is read from the storage device 1 or another storage device under control of the outside or the power consumption totaling means 5 that is stored in advance in the storage device 1 or another storage device, It may be configured to be input to the consumption calculation means 4 or may be directly input to the power consumption calculation means 4 by an input operation by an operator or the like.

  Next, in step # 21, it is specified to which of the first segments determined the city gas consumption amount G is determined for the group specified in the first process described above.

  Next, at step # 22, among the calculation formulas preset for each second segment, the calculation formula of the second segment corresponding to the first segment identified at step # 21 is selected, and calculation is performed at step # 20. The power consumption E is calculated by substituting the received or received city gas consumption G into the selected calculation formula.

  The processes of steps # 20 to # 22 are executed for all consumers to be estimated. The power consumption totaling means 5 totals the power consumption E calculated by the power consumption calculation means 4 for each consumer, multiplies the maximum value of the power consumption, and calculates all the consumers to be estimated. Calculate the total power consumption of the month. When the process is performed from January to December, the total annual power consumption is calculated.

  Next, the calculation formula used in step # 22 will be described. In this embodiment, a calculation formula derived in advance for each second segment by the calculation formula deriving means 6 is used. As calculation formulas, use of the following three types of calculation formulas is assumed. The first calculation equation is a linear expression shown in the following equation 2, the second calculation equation is a simple regression equation shown in the following equation 3, and the third calculation equation is the following equation 4 It is a multiple regression equation shown.

(Equation 2) E = a1 × G + a0
(Equation 3) E = b1 × G + b0
(Equation 4) E = c1 × G + c2 × F2 + ······ + ck × Fk + c0

  E in the numbers 2 to 4 is the power consumption E, and G is the city gas consumption G. In the simple regression equation of Equation 3, the power consumption E is a target variable, the city gas consumption G is an explanatory variable, and the number of explanatory variables is one. In the multiple regression equation of Equation 4, the power consumption E is the objective variable, the city gas consumption G is the explanatory variable, and besides the city gas consumption G, it also relates to (k-1) attributes of consumers Explanatory variables F2 to Fk are used, and the number of explanatory variables is k (k is an integer of 2 or more). The explanatory variables F2 to Fk may be all or part of the attributes used to classify the group described above, and may further add an attribute not used to classify the group. Therefore, the explanatory variables F2 to Fk may be discrete variables. Each of a1, b1 and c1 is a coefficient of the city gas consumption G, and a0, b0 and c0 are respective constants. c2 to ck are coefficients of the explanatory variables F2 to Fk.

  Hereinafter, a method of deriving each of the calculation equations shown in Equations 2 to 4 by the calculation equation deriving means 6 will be described. When the calculation formula is a linear expression shown in Equation 2, the calculation formula is derived for each second segment and for each group. Specifically, in the above-described first process, the first threshold values TG1j to TG (m-1) j corresponding to the second threshold values TE1 to TE (m-1) determined for each group are divided into second segments. In other words, two second thresholds TEi and second thresholds TE (i + 1) defining the second segment, and two second thresholds TGij and first thresholds TG (i + 1) defining the corresponding first segments Substituting j into the first equation, the coefficient a1 and the constant a0 are calculated for each second segment and each group. In other words, the calculation equation shown in Equation 2 is represented by a broken line shown by a broken line in FIG. In the case where the linear expression shown in equation 2 is used as a calculation equation in step # 22 of the second process, it corresponds to the group identified in step # 20 and the first segment identified in step # 21. The calculation formula of the second segment will be selected and used.

  In the case where the calculation formula is a single regression formula shown in Equation 3, the calculation formula is derived by performing a single regression analysis by grouping all the groups for each second segment. Specifically, for the i-th second segment defined by the second threshold TE (i-1) and the second threshold TEi, all sample values present in the i-th second segment are The power consumption E and the city gas consumption G are used to perform known single regression analysis by, for example, the least squares error method. If the number of samples present in the corresponding i-th first segment among the samples present in the i-th second segment is sufficient for the regression analysis, the i-th The above single regression analysis may be performed using the power consumption E of the sample value and the city gas consumption G which are present in the two segments and also in the corresponding i-th first segment. As the method of selecting sample values for securing the number of samples necessary for single regression analysis, various methods other than the above can be considered.

  When the calculation formula is a multiple regression formula shown in Equation 4, the calculation formula is derived by performing multiple regression analysis on all the groups for every second segment. Specifically, for the i-th second segment defined by the second threshold TE (i-1) and the second threshold TEi, all sample values present in the i-th second segment are Using the power consumption E, the city gas consumption G, and the values of the explanatory variables F2 to Fk according to the attributes of the consumer of the sample value, for example, the well-known multiple regression analysis is performed by the least square error method. If the number of samples present in the corresponding i-th first segment among the samples present in the i-th second segment is sufficient for the regression analysis, the i-th The above-described multiple regression analysis may be performed using the power consumption E of the sample value and the city gas consumption G which are present in two segments and also in the corresponding i-th first segment. As a method of selecting sample values for securing the number of samples necessary for multiple regression analysis, various methods other than the above can be considered. When the multiple regression equation shown in equation 4 is used as a calculation equation in step # 22 of the second process, when the city gas consumption of one consumer is accepted together with the group number in step # 20, The values of explanatory variables F2 to Fk related to the attributes of the consumer are also received at the same time, and in step # 22, the explanatory variables F2 to Fk are calculated together with the city gas consumption amount G calculated or received in step # 20. The power consumption E is calculated by substituting the equation into the equation. When the group number and the values of the explanatory variables F2 to Fk are uniquely associated, a table indicating the association is separately prepared, and the values of the explanatory variables F2 to Fk are obtained from the group number accepted by the table. May be read out.

  Among the calculation formulas shown in the above-described Eqs. 2 to 4, the multiple regression equation shown in Eq. 4 is the best in terms of the estimation accuracy of the power consumption E, and the linear equation shown in Eq. In the simple regression equation, there is no big difference. However, when the positive correlation between the power consumption amount E of the sample value to be used and the city gas consumption amount G is strong, there is a possibility that high estimation accuracy can be obtained even by the calculation formula shown in Equation 2 or Equation 3.

  Next, with respect to the estimation accuracy by the device 1 of the present invention, the difference between the difference due to the presence or absence of the first processing, the difference due to the presence or absence of the group classification, and the difference due to the difference between the calculation formulas shown in Equations And will be described with reference to FIG.

  If the first process is not performed, the power consumption E and the city gas consumption G are not divided into segments, so the calculation formula used in the second process is a single regression equation shown in Eq. 3 or Eq. 4 It becomes the multiple regression equation shown. When the first process is performed, the presence or absence of the classification of the group is distinguished. If the grouping of groups is performed, the first threshold value TGij (i = 1 to m-1) is derived for each group j because the first process is performed for each group when the grouping of groups is performed. Also in the processing, the specification of the first segment in step # 21 and the selection of the calculation formula in the case of using the calculation formula shown in Formula 2 are performed according to the group j. On the other hand, when the classification of the group is not performed, the first threshold value TGij (i = 1 to m-1) is derived in one way regardless of the group because the first process is not executed for each group. Also in the second process, the specification of the first segment in step # 21 and the selection of the calculation formula in the case of using the calculation formula shown in Formula 2 are performed regardless of the group.

Comparison examination was performed about eight cases of the following cases AH. Cases A and B are comparative examples to the device 1 of the present invention, and cases C to H are embodiments of the device 1 of the present invention.
Case A: first process = no, calculation formula = single regression formula shown in equation 3 Case B: first process = no, calculation formula = multiple regression formula shown in formula 4 Case C: first process = yes, Group classification = No, Calculation formula = Primary expression shown in Equation 2. Case D: First processing = Yes, Group classification = No, Calculation formula = Single regression formula shown in Expression 3. Case E: First process = Yes, Group classification = None, Calculation formula = Multiple regression shown in Formula 4 Case F: First processing = Yes, Group classification = Yes, Formula = Linear expression shown in Formula 2 Case G: First process = Yes, Group classification = presence, calculation formula = single regression formula shown in equation 3 Case H: first processing = presence, group classification = presence, calculation formula = multiple regression formula shown in formula 4

  For cases A to H, the yearly total electric power consumption (estimated value) and total of monthly from April to December of the year and January to March of the next year for all consumers targeted for estimation The total power consumption (estimated value) of is calculated, and from the true value of each total power consumption and the estimated value, the error of each estimated value is calculated by the following equation 5. The calculated errors (rounded off after the decimal point) are listed in FIG. The number of samples used in this comparative study is 8000, the number of segments in cases C to H is 5, the number of groups in cases F to H is 20, and the number of explanatory variables of the multiple regression of cases B, E and H (city The gas consumption G is 4). Further, as the first classification standard of the first processing, the hit rate Rijij70% was adopted.

  (Equation 5) Error = | true value−estimated value | / true value × 100 (%)

  Furthermore, the error improvement effect of performing the first processing (case A to case D, case B to case E), the error improvement effect of performing group classification (case C to case F, case D to case G to case E, case E To case H), and the linear equation of equation 2 to a simple regression equation of equation 3 (hereinafter referred to as “single regression equation” as appropriate) error improvement effects (case C to case D, case F to case) G) The error improvement effect (case A to case B, case D to case E, changing the single regression equation of equation 3 to the multiple regression equation of equation 4 (hereinafter referred to as “multiple regression equation” as appropriate) Case G to case H), error improvement effect of performing the first process and group classification (case A to case G, case B to case H), error improvement effect of performing the first process and multiple regression formula ( Case A to Case E), group minutes Error improvement effect (case D to case H) of performing multiple regression formula and error improvement effect (case A to case H) of performing the first processing, group classification, and multiple regression formula, List on 8 In addition, the error improvement effect was evaluated by the error improvement rate (rounded off after the decimal point) shown to the following number 6. In Equation 6, the error X is the error in the case before improvement, and the error Y is the error in the case after improvement. For example, in the case of the error improvement effect (case A to case H) of performing the first processing and group classification and multiple regression, the error X is the error of case A and the error Y is the error of case H.

  (Equation 6) error improvement rate = (error X−error Y) / error X × 100 (%)

  As shown in FIG. 7, in the case A (comparative example) using the single regression equation without performing the first processing and the group classification, the monthly error is 29% to 36%, and the average monthly error is 32%. The annual error is 30%, and the error exceeds 30% with high probability. As the error in estimating the monthly total power consumption and the total power consumption yearly, an error exceeding 30% is not preferable. However, in Case D, where the first process is added to Case A, the monthly error is reduced to 25% to 32%, the average monthly error is reduced to 28%, and the annual error is reduced to 26%. The average of the error improvement rate of is 12%. On the other hand, in case B (comparative example) in which multiple regression is performed without performing the first process on case A, the monthly error is 25% to 31%, the average monthly error is 28%, and The errors each decrease to 26%, and the average monthly error improvement rate is 11%. Here, it can be seen that the error improvement effect by the addition of the first process to the case A is substantially equal to the error improvement effect in the multiple regression formula without the first process, but it is slightly larger. In each of Case B and Case D, the monthly error exceeds 30% in the cold period of January to March when the city gas consumption is high, but the error is improved by 11% or more as a whole. Furthermore, in Case E where Case 1 was subjected to the first treatment and multiple regression, the monthly error was 21% to 29%, the average monthly error was 25%, and the annual error was 22%. It is decreasing, and the monthly error is less than 30% throughout the January-December year. As described above, according to the results of Case D and Case E, the error improvement rate by the addition of the first process is 11% or 12% when the calculation formula is any of the single regression equation and the multiple regression equation. By performing both of the multiple regression formulas, the error improvement rate becomes 20%, and a further error improvement effect can be expected.

  Next, the error improvement effect by performing group classification is examined. In case C of the linear expression in which the calculation formula without grouping is shown in Equation 2 where the first process is added to case A, the monthly error is 26% to 32%, and the average monthly error is 29%, While the annual error is 26%, in Case F where the group classification is added to Case C, the monthly error is 24% to 29%, the average monthly error is 26%, and the annual error is Each has been reduced to 23%, a clear error improvement effect has been seen from Case C to Case F, and the average monthly error improvement rate is 10%. In addition, in case D of the simple regression equation in which the calculation formula without grouping is added to the first process with respect to case A is shown in Equation 3, the monthly error is 25% to 32%, and the average monthly error is 28 %, The annual error is 26%, whereas in Case G where the group classification is added to Case D, the monthly error is 23% to 28%, and the average monthly error is 25%, yearly The errors are respectively reduced to 22%, a clear error improvement effect is seen from case D to case G, and the average monthly error improvement rate is 10%. Furthermore, in Case E of the multiple regression equation in which the calculation formula without grouping is added to the first process with respect to Case A is shown in Equation 4, the monthly error is 21% to 29%, and the average monthly error is 25. %, Annual error is 22%, in case H where group classification is added to case E, monthly error is 20% -25%, average of monthly error is 22%, annual The errors are respectively reduced to 18%, a clear error improvement effect is seen from case E to case H, and the average monthly error improvement rate is 13%. From the above, the error improvement effect by the classification of the group is clear whether the calculation formula is any of the linear expression shown in Equation 2, the single regression equation shown in Equation 3, and the multiple regression equation shown in Equation 4. In particular, in the case of the multiple regression equation shown in Equation 4, a more remarkable error improvement effect can be seen.

  Next, the error improvement effect by performing both the first processing and the group classification is examined. As described above, although the first processing and the group classification each have an error improvement effect even if they are independent, by simultaneously performing both, the error improvement effects of both are added and exhibited. Assuming that the error improvement rate in the first process is A (%), the error improvement rate in the group classification is B (%), and the error improvement rate in both the first process and the group classification is C (%), the error improvement rate C is , Is represented by the following number 7. However, since group classification presupposes 1st process, the error improvement rate by group classification is based on the error at the time of performing 1st processing.

  (Equation 7) C = A + B−A × B / 100 (%)

  From the error improvement rate shown in FIG. 8, the average of the monthly error improvement rates by the first process from case A to case D is 12%, and the monthly error improvement rates by the first process from case B to case E are The average is 11%. In addition, the average of the monthly error improvement rate by the group classification from Case D to Case G is 10%, and the average of the monthly error improvement rate by the group classification from Case E to Case H is 13%. On the other hand, the average of the monthly error improvement rate by the first process from Case A to Case G and the group classification is 20%, and the monthly error improvement by the first process from Case B to Case H and the group classification It can be seen that the average of the rates is 22%, and the relationship shown in the above equation 7 is satisfied.

  Next, Case E in which multiple regression expression is made without performing group classification on Case D and Case G in which group classification is added to Case D with the calculation formula being unchanged are compared. In Case E, the monthly error is 21% to 29%, the average monthly error is 25%, and the annual error is 22%, whereas in Case G, the monthly error is 23% to 28%, The average monthly error is 25%, and the annual error is 22%, and the errors in both cases are similar. The average monthly error improvement rate from Case D to Case E is 10%, while the average monthly error improvement rate from Case D to Case G is 10%, and the error improvement effect in both cases is approximately It is equivalent. Furthermore, the average monthly error improvement rate by simultaneously performing the group classification from Case D to Case H and the multiple regression formula is 22%.

  From the above, in addition to the first processing, there is an error improvement effect by performing multiple regression expression without performing group classification, and an error improvement effect by performing group classification with single regression expression shown in equation 3. Almost identical, in addition to the first processing, group classification is performed, and further, the error improvement effect by multiple regression expression is more effective than simply adding the above two error improvement effects I understand that.

  Next, the error improvement effect by making a single regression formula is examined. In the case where there is no group classification, in the case C of the linear equation shown in the equation 2, the monthly error is 26% to 32%, the average monthly error is 29%, and the annual error is 26%, In case D of the simple regression equation whose calculation formula is shown in Equation 3, the monthly error is 25% to 32%, the average monthly error is 28%, and the annual error is 26%. The average monthly error improvement rate by the simple regression formula is only 2%, and with the May and annual errors, the error improvement rate is negative, and Case C has a smaller error. In Case C and Case D, no significant difference is found in the estimation accuracy. In the case where there is a group classification, in case F of the linear equation whose calculation formula is shown in Equation 2, the monthly error is 24% to 29%, the average monthly error is 26%, and the annual error is 23%, In case G of the single regression equation whose calculation formula is shown in Equation 3, the monthly error is 23% to 28%, the average monthly error is 25%, and the annual error is 22%. The average monthly error improvement rate by simple regression formula is only 2%, and the error improvement rate (assessed after the decimal point) is negative for errors in February, April, July and October. In Case F, the error is smaller, and in Case F and Case G, a large difference in estimation accuracy is not recognized.

  As described above, the error improvement effect by the first processing, the error improvement effect by the group classification, and further, the error improvement effect by the multiple regression formulaization were respectively confirmed.

  Another embodiment will be described below.

  <1> In the above embodiment, the device 1 of the present invention includes the storage unit 2, the first threshold calculation unit 3, the power consumption calculation unit 4, the power consumption totaling unit 5, and the calculation formula derivation unit 6. In the case where the present invention device 1 is used to associate the segment between the power consumption and the city gas consumption, the present invention device 1 does not necessarily have to use the first threshold calculation means 3 and the power. It is not necessary to provide the consumption calculation means 4, the power consumption totaling means 5, and the calculation formula derivation means 6.

  Further, in the case of separately totaling the power consumption per consumer calculated by the power consumption calculation means 4, the power consumption totaling means 5 may be provided separately from the device 1 of the present invention. For example, it is assumed that the power consumption per consumer calculated by the power consumption calculation unit 4 is stored in the storage unit 2.

  Furthermore, in the case of fixed use without updating the calculation formula of the power consumption calculation means 4, the calculation formula derivation means 6 is provided separately from the device 1 of the present invention, and the calculation formula derived in advance is provided. It is also preferable to set the power consumption calculation means 4 in the following manner.

  Furthermore, the present invention device 1 may not have the first threshold value calculating means 3. However, in this case, the first threshold value TGij preset by the first threshold value calculation unit 3 is stored in the storage unit usable by the storage unit 2 or the power consumption calculation unit 4 and used in the second process. It is preferable to configure as possible.

  <2> In the description of the first process of the above embodiment, the power consumption and the city gas consumption are treated as the power consumption E and city gas consumption G normalized respectively, but without normalization The actual consumption amounts may be processed.

  Further, in the description of the first process of the above embodiment, the first and second classification criteria are satisfied in step # 12 with respect to the first first segment for which the city gas consumption amount G is 0 to TG1 j. Since it is not determined whether or not it is determined, it is unclear whether the first classification criterion is satisfied in the first first segment. Therefore, in the first first segment, a step of determining whether or not the first classification criterion is satisfied is additionally provided, and in the case where the first classification criterion is not satisfied in the judgment, the second In step # 12 for the first segment, the second classification criterion is satisfied for the second first segment, provided that the first and second first segments simultaneously satisfy the first classification criterion. The first threshold value TG1 j may be set (action 1). Furthermore, in this case, when the first and second first segments can not simultaneously set the first threshold TG1j that satisfies the first reference, the second threshold TE1 set in the pre-processing of the first processing is set. Then, the first process may be performed again (action 2).

  Furthermore, in the case of performing the treatment 1 and the treatment 2, since the number N1 j of hits in the second first segment decreases from the initial maximum value, the decrease is made to the third to m first segments. The first threshold value TG2j to the first threshold value TG (m-1) j may also be set again so that the number N1j of the number of hits in the second first segment is not significantly reduced. Therefore, in the first first segment, if the first classification criterion is not satisfied, the treatment satisfying the second classification criterion in step # 12 is a provisional treatment.

  Furthermore, in the description of the first process of the above embodiment, setting of the first threshold value TGij is performed in order from the m-th first segment to the second first segment, but conversely, the first first segment is set To the (m-1) -th first segment may be performed in order. However, step # 12 causes the first threshold value TGij to be increased stepwise in predetermined small steps. Furthermore, in this case, since the possibility that the first classification criterion is not satisfied in the above-described first first segment occurs in the m-th first segment, the same treatment as the above-mentioned treatment 1 or treatment 2 may be performed. Conceivable. Furthermore, regardless of the setting order of the first threshold value TGij, at least one of the first and m-th first segments is used as a dummy first segment that allows the first classification criterion not to be satisfied. The dummy second segment corresponding to may be set in the pre-processing of the first processing.

  Furthermore, in the description of the first process of the above embodiment, it is assumed that the setting of the first threshold TGij is performed for each group j, but a stronger positive correlation between the power consumption E and the city gas consumption G In some cases or when it is possible to use a multiple regression equation with a higher error improvement effect as a calculation equation, the above-mentioned group classification is not necessarily essential. However, it is preferable to perform group classification as much as possible because further error improvement effects can be surely expected by group classification.

  <3> In the above embodiment, one month is assumed as the first period and the second period, and it is assumed that the first period and the second period are the same month, but sample values other than the one month The first period and the second period are limited to one month if there is data of the combination of the consumption of the first energy in the first period and the consumption of the second energy in the second period. Furthermore, it may be a short period (for example, one week, one day, one hour, etc.), and the first period and the second period may not be the same period. For example, if the correlation between the consumption of the first energy and the consumption of the second energy in the second period is weak in a specific period of one year, it may be different from the consumption of the second energy in the specific period. When the correlation of the consumption of the first energy in the period is stronger, it is also preferable to associate the consumption of the second energy in the specific period with the consumption of the first energy in the other period.

  Furthermore, although the case where city gas and electric power were assumed as the 1st and 2nd energy which the present invention device 1 deals with, and power consumption was estimated from city gas consumption was explained, the present invention device 1 conversely, It can be applied to estimate city gas consumption from electricity consumption. Furthermore, one or both of the first and second energy may be an energy species other than city gas and power. In this case, the first and second energies do not necessarily have to be positively correlated, but may be negatively correlated. That is, it is necessary for the first and second energies to have a positive or negative correlation.

  <4> In the above embodiment, the estimation accuracy by the device 1 of the present invention has been described with reference to FIGS. 7 and 8. However, the number of samples, the number of segments, the number of groups, the explanatory variables of multiple regression equation used in the description The number, the lower limit value of the accuracy of the first classification standard, and the like are examples, and the device 1 of the present invention is not limited to the specific example.

  <5> In the above embodiment, the device 1 of the present invention is assumed to be used as a heterogeneous energy consumption estimation device, but in addition to the estimation of consumption between dissimilar energy, two different events are used. When there is a positive or negative correlation between the distribution of one of the first characteristic values and the distribution of the other of the second characteristic values in the set of individuals each exhibiting a predetermined characteristic, one of the first characteristic values It is also possible to use it as a heterogeneous characteristic estimating device which estimates the second characteristic value of. That is, the device 1 of the present invention as the different energy consumption estimation device of the above embodiment can be regarded as one aspect of the different characteristic estimation device.

  In this case, in the first processing by the first threshold value calculation means 3, one or more first threshold values which are boundary values for dividing the distribution range of the first characteristic value into a plurality of first segments, the distribution range of the second characteristic value Are respectively corresponded to one or more second threshold values set in advance, which are boundary values for dividing into a plurality of second segments, and the processing is respectively calculated. Here, as the first and second classification criteria, those described in the first process of the embodiment can be used as they are. Group classification can also be handled in the same manner as in the above embodiment by performing classification on the basis of one or more attributes of the individual, and setting of the first threshold is also performed on a group basis. As an example, it is assumed that one of the first and second characteristic values is energy consumption such as the above-mentioned city gas or electric power, and the other is consumption other than energy, for example, specific household expenses such as food expenses etc. Ru.

  The heterogeneous energy consumption estimation apparatus according to the present invention can be used to estimate the consumption of another energy from the consumption of one energy of a consumer who consumes multiple types of energy such as power.

1: Heterogeneous energy consumption estimation device 2: Storage means 3: First threshold value calculation means 4: Power consumption calculation means 5: Power consumption amount collection means 6: Calculation formula derivation means

Claims (15)

The consumption of the first energy in the first period of each consumer and the same or different second period of the first period in a group of consumers consuming different first energy and second energy. From the known consumption of the first energy of the first period based on the known sample data in which the pairs of consumption of the second energy are individual sample values, the unknown second of the second energy period is used. A heterogeneous energy consumption estimation device for estimating energy consumption, comprising:
Storage means for storing the sample data;
Distribution of consumption of the second energy in the second period, one or more first threshold values being boundary values dividing the distribution range of consumption of the first energy in the first period into a plurality of first segments First threshold calculation means for calculating each corresponding to one or more preset second thresholds which are boundary values dividing the range into a plurality of second segments, respectively;
The first threshold value calculation means, for all of the first segments or for each of preselected ones, corresponds to the first one of the total number of the sample values belonging to one first segment. The different energy consumption amount estimation apparatus, wherein the first threshold is calculated such that the hit rate, which is a ratio of the number of sample values belonging to two segments, is equal to or more than a predetermined value.   One of the first threshold value calculation means corresponds to one of the first segments in a range in which the hit ratio is equal to or more than the predetermined value with respect to each of the previously selected portions of the first segments. The heterogeneous energy consumption amount estimation apparatus according to claim 1, wherein the first threshold is calculated such that the number of the sample values belonging to both of the second segments is the maximum. . The assembly is divided into a plurality of groups based on one or more attributes of the consumer,
The storage means stores the sample data so as to distinguish which of the plurality of groups the individual sample values belong to;
The one or more second thresholds are commonly set through the plurality of groups.
The different energy consumption amount estimation apparatus according to claim 1 or 2, wherein the first threshold calculation means calculates the first threshold corresponding to the second threshold separately for each of the groups. The consumption amount of the first energy of the first period of one of the consumers selected from the aggregate is received as an input, and the calculation formula set for each of the second segments is used for the second period of the second period. A second energy consumption calculation unit that calculates the consumption of the second energy;
The second energy consumption calculation means compares the consumption of the first energy received as an input with the first threshold to determine which of the plurality of first segments the energy consumption belongs to, The amount of consumption of the second energy in the second period is calculated using the calculation formula set for the second segment corresponding to the determined first segment. The heterogeneous energy consumption estimation device according to 2. The amount of consumption of the first energy in the first period of the consumer selected from the aggregate and identification information of the group to which the consumer belongs are received as inputs, and are set for each of the second segments. A second energy consumption amount calculation unit configured to calculate the consumption amount of the second energy in the second period using the calculated expression formula;
The second energy consumption amount calculation means compares the consumption amount of the first energy received as an input with the first threshold value calculated for the group of the identification information received as an input. It is determined which of the first segments it belongs to, and using the calculation formula set for the second segment corresponding to the determined first segment, the second of the second period The different energy consumption estimation apparatus according to claim 3, wherein the consumption of energy is calculated. From the known consumption of the first energy of the first period per consumer in a collection of consumers consuming different first energy and second energy from each other, the same or different first energy as the first period What is claimed is: 1. A heterogeneous energy consumption estimation device for estimating the consumption of the second energy unknown for two periods, comprising:
Storage means for storing one or more first threshold values which are boundary values for dividing the distribution range of consumption of the first energy in the first period into a plurality of first segments;
The consumption of the first energy in the first period of the one consumer selected from the assembly is received as an input, and the consumption of the second energy in the second period is calculated using a predetermined calculation formula. And second energy consumption calculating means for calculating
The pair of the known consumption of the first energy in the first period and the consumption of the second energy in the second period of each consumer in the assembly is taken as an individual sample value;
The first threshold corresponds to one or more second thresholds, which are boundary values dividing the distribution range of the consumption of the second energy of the second period into a plurality of second segments, Of the number of sample values belonging to the corresponding one second segment among the total number of sample values belonging to one of the first segments, for each of all or a portion selected in advance. It is set so that a certain hit rate is equal to or more than a predetermined value,
The calculation formula is set for each of the second segments,
The second energy consumption calculation means compares the consumption of the first energy received as an input with the first threshold to determine which of the plurality of first segments the energy consumption belongs to, The consumption amount of the second energy in the second period is calculated using the calculation formula set for the second segment corresponding to the determined first segment. Estimator. From the known consumption of the first energy of the first period per consumer in a collection of consumers consuming different first energy and second energy from each other, the same or different first energy as the first period What is claimed is: 1. A heterogeneous energy consumption estimation device for estimating the consumption of the second energy unknown for two periods, comprising:
The assembly is divided into a plurality of groups based on one or more attributes of the consumer,
One or more first thresholds, which are boundary values for dividing the distribution range of the consumption of the first energy in the first period into a plurality of first segments, and each of the consumers belongs to any of the plurality of groups Storage means for storing information capable of identifying
The consumption amount of the first energy of the first period of the consumer selected from the aggregate and the identification information of the group to which the consumer belongs are received as input, using a predetermined calculation formula, A second energy consumption amount calculation unit configured to calculate the consumption amount of the second energy in the second period;
The pair of the known consumption of the first energy in the first period and the consumption of the second energy in the second period of each consumer in the assembly is taken as an individual sample value;
The first threshold corresponds to one or more second thresholds, which are boundary values dividing the distribution range of the consumption of the second energy of the second period into a plurality of second segments, Of the number of sample values belonging to the corresponding one second segment among the total number of sample values belonging to one of the first segments, for each of all or a portion selected in advance. It is set according to the above-mentioned group so that a certain hit rate will be above a predetermined value
The calculation formula is set for each of the second segments,
The second energy consumption amount calculation means compares the consumption amount of the first energy received as an input with the first threshold set for the group of the identification information received as an input. It is determined which of the first segments it belongs to, and using the calculation formula set for the second segment corresponding to the determined first segment, the second of the second period What is claimed is: 1. A heterogeneous energy consumption estimation device for calculating energy consumption.   One of the first corresponding to the one first segment in a range in which the hit ratio is equal to or more than the predetermined value with respect to each of the preselected portions of the first segment as the first threshold. The different energy consumption amount estimation apparatus according to claim 6 or 7, wherein the number of sample values belonging to both of two segments is set so as to maximize the number of the sample values.   A regression equation set for each of the second segments in which the calculation formula uses the consumption of the second energy in the second period as a target variable and the consumption of the first energy in the first period as an explanatory variable The heterogeneous energy consumption estimation apparatus according to any one of claims 4 to 8, wherein:   10. The system according to claim 9, wherein the regression equation is a multiple regression equation further including, as an explanatory variable, at least one of the one or more attributes of the consumer selected from the group. Heterogeneous energy consumption estimation device.   11. The heterogeneous energy consumption amount estimation apparatus according to claim 9, wherein the regression equation is a regression equation derived by regression analysis using the sample values belonging to the corresponding second segment.   The consumers are selected one by one from the assembly, and the consumption amount of the first energy of the consumer during the first period is input to the second energy consumption calculation means, and the second energy consumption is obtained. The second energy consumption totaling means for totaling the consumption amount of the second energy in the second period sequentially outputted from the amount calculation means is provided, according to any one of claims 4 to 11, Heterogeneous energy consumption estimation device. Known sample data in which a pair of a first characteristic value of one of the two events and a second characteristic value of another of the two events in an aggregate of individuals exhibiting predetermined characteristics for two different events is an individual sample value A heterogeneous characteristic estimation apparatus for estimating the unknown second characteristic value from the known first characteristic value based on
Storage means for storing the sample data;
The boundary value dividing the distribution range of the second characteristic value into a plurality of second segments, one or more first threshold values being boundary values dividing the distribution range of the first characteristic value into a plurality of first segments First threshold calculation means for respectively calculating one or more second threshold values set in advance,
The first threshold value calculation means, for all of the first segments or for each of preselected ones, corresponds to the first one of the total number of the sample values belonging to one first segment. The first and second threshold values are respectively calculated such that the hit rate, which is a ratio of the number of sample values belonging to two segments, is equal to or greater than a predetermined value.   One of the first threshold value calculation means corresponds to one of the first segments in a range in which the hit ratio is equal to or more than the predetermined value with respect to each of the previously selected portions of the first segments. The different characteristic estimation apparatus according to claim 13, wherein the first threshold is calculated such that the number of the sample values belonging to both of the second segments is the maximum. The assembly is divided into a plurality of groups based on one or more attributes of the individual,
The storage means stores the sample data so as to distinguish which of the plurality of groups the individual sample values belong to;
The one or more second thresholds are commonly set through the plurality of groups.
The heterogeneous characteristic estimation apparatus according to claim 13 or 14, wherein the first threshold calculation means calculates the first threshold corresponding to the second threshold separately for each of the groups.