Model generation device, prediction device, model generation method, and prediction method

Abstract

To improve the accuracy of predicting the amount of power demanded in a facility.SOLUTION: A model generation apparatus (10) is configured to: derive, for time-series data indicating temporal transition of past results values of the amount of power demanded in a facility, a partial autocorrelation function of the past results values corresponding to a plurality of lags, by setting the lags for the past results values; select a special lag from among the lags on the basis of the partial autocorrelation function; derive new feature quantity by setting the special lag to the past results values; and generate, based on the time-series data with the new feature quantity added thereto, a prediction model for predicting the amount of power demanded in the facility by machine learning.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] One aspect of the present invention relates to a model generation device that generates a prediction model for predicting the required power consumption in a facility.

Background Art

[0002] In recent years, machine learning has been used to predict various types of energy consumption in facilities. For example, Patent Document 1 below discloses a technique for predicting the heat load in a heat storage utilization system using a neural network.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] One aspect of the present invention aims to improve the prediction accuracy of the required power consumption in a facility.

Means for Solving the Problems

[0005] In order to solve the above problems, a model generation device according to one aspect of the present invention is a model generation device that generates a prediction model for predicting the required power consumption in a facility, and the model generation device sets a plurality of lags with respect to the actual value for time series data indicating the time transition of the actual value of the required power consumption, and derives the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags, selects a special lag from among the plurality of lags based on the partial autocorrelation coefficient, and sets the special lag with respect to the actual value to derive a new feature amount, and generates the prediction model by machine learning based on the time series data with the new feature amount added.

[0006] Also, a prediction device according to one aspect of the present invention is a prediction device that derives a predicted value of the required power consumption by using, in a prediction phase, a prediction model for predicting the required power consumption in a facility, which was generated in advance in a learning phase. In the learning phase, for time series data in the learning phase that shows the temporal change of the actual value of the required power consumption, by setting a plurality of lags with respect to the actual value, the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags is derived. Based on the partial autocorrelation coefficient, a special lag is selected from among the plurality of lags. By setting the special lag with respect to the actual value, a new feature amount is derived. Based on the time series data with the new feature amount added, the prediction model is generated by machine learning. The prediction device sets the special lag with respect to the actual value in the time series data in the prediction phase that shows the temporal change of the actual value, thereby deriving a new feature amount in the prediction phase, and outputs the predicted value to the prediction model based on the time series data in the prediction phase with the new feature amount added in the prediction phase.

[0007] Also, a model generation method according to one aspect of the present invention is a model generation method for generating a prediction model for predicting the required power consumption in a facility, including: a step of deriving the partial autocorrelation coefficient of the actual value corresponding to each of a plurality of lags by setting the plurality of lags with respect to the actual value for time series data showing the temporal change of the actual value of the required power consumption; a step of selecting a special lag from among the plurality of lags based on the partial autocorrelation coefficient; a step of deriving a new feature amount by setting the special lag with respect to the actual value; and a step of generating the prediction model by machine learning based on the time series data with the new feature amount added.

[0008] Moreover, the prediction method according to one aspect of the present invention is a prediction method for deriving a predicted value of the required power consumption in a facility by using, in a prediction phase, a prediction model for predicting the required power consumption in the facility that has been generated in advance in a learning phase. In the learning phase, for the time-series data in the learning phase indicating the temporal change of the actual value of the required power consumption, a plurality of lags with respect to the actual value are set, and the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags is derived. Based on the partial autocorrelation coefficient, a special lag is selected from among the plurality of lags, and a new feature amount is derived by setting the special lag with respect to the actual value. Based on the time-series data with the new feature amount added, the prediction model is generated by machine learning. The prediction method includes a step of deriving a new feature amount in the prediction phase by setting the special lag with respect to the actual value in the time-series data in the prediction phase indicating the temporal change of the actual value, and a step of causing the prediction model to output the predicted value based on the time-series data in the prediction phase with the new feature amount added in the prediction phase.

Effect of the Invention

[0009] According to one aspect of the present invention, the prediction accuracy of the required power consumption in a facility can be improved.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Best Mode for Carrying Out the Invention

[0011] 〔Embodiment 1〕 Embodiment 1 will be described below. For convenience of explanation, components having the same functions as those described in Embodiment 1 are denoted by the same reference numerals in the following embodiments, and the description thereof will not be repeated. For the sake of simplicity, descriptions of known technical matters will also be omitted as appropriate. Each component and each numerical value described in this specification are merely illustrative examples as long as there is no particular contradiction. Therefore, for example, the positional relationship and connection relationship of each component are not limited to the examples in each figure as long as there is no particular contradiction. Also, note that the correspondence between the dates and days of the week in the examples of each figure does not necessarily match those of an actual calendar. In this specification, as long as there is no particular contradiction, the notation "A~B" for two numbers A and B represents "A or more and B or less".

[0012] (Overview of Information Processing Apparatus 1) FIG. 1 is a block diagram illustrating the configuration of the main part of the information processing apparatus 1. The information processing apparatus 1 may include a control device 9 and a storage unit 90. The control device 9 comprehensively controls each part of the information processing apparatus 1. The control device 9 may include a model generation device 10 and a prediction device 20.

[0013] As will be described later, the model generation device 10 generates a prediction model for predicting the required power consumption in a facility by machine learning. Therefore, the model generation device 10 may also be referred to as a learning device. The prediction device 20 uses the prediction model previously generated by the model generation device 10 to derive a predicted value of the required power consumption in the above-mentioned facility.

[0014] The facility in one aspect of the present invention may be, for example, a facility in which fluctuations in the required power consumption can occur according to at least one of a time zone, a day of the week, and the employment status of employees in the facility. Examples of the facility in one aspect of the present invention include industrial facilities such as factories, commercial facilities such as shopping malls, and medical facilities such as hospitals.

[0015] The storage unit 90 stores various types of data and programs used in the processing of the control device 9. In the example of FIG. 1, it is assumed that the performance data and the working day data are stored in the storage unit 90. In this specification, the performance data generally refers to data (more specifically, data structures) indicating the actual values of the required power consumption in the facility recorded at a predetermined time resolution. In this specification, unless otherwise inconsistent, the term "required power consumption" refers to the actual value of the required power consumption.

[0016] Since the performance data is data in which a time point and the required power consumption are associated, it is an example of time-series data showing the time transition of the required power consumption. In the example of Embodiment 1, the time resolution in the performance data is 1 minute. In Embodiment 1, it is assumed that the time points in the data with a time resolution of 1 minute are expressed in the format of "year / month / day hour / minute" ("yyyy / mm / dd / hh / mm" format). The time point may be used as a key in the data structure.

[0017] In this specification, the working day data generally refers to data indicating the daily working status in the facility. As an example, the working day data may be data indicating whether a certain day is a working day of the facility. Since the working day data is data in which a date and the working status are associated, it is an example of time-series data showing the daily transition of the working status. In the example of Embodiment 1, the time resolution in the working day data before the following pre-interpolation is performed is 1 day. In Embodiment 1, it is assumed that the dates in the data with a time resolution of 1 day are expressed in the format of "year / month / day" ("yyyy / mm / dd" format).

[0018] (Learning phase) Hereinafter, an example of the operation of the information processing apparatus 1 will be described. The processing of the information processing apparatus 1 is roughly classified into processing in the learning phase (a series of processing of the model generation apparatus 10) and processing in the prediction phase (a series of processing of the prediction apparatus 20). First, an example of the learning phase will be described.

[0019] As shown in FIG. 1, the model generation device 10 may include a data acquisition unit 11, a preprocessing unit 12, a learning unit 13, and an evaluation unit 14. The data acquisition unit 11 may also be referred to as a learning-time data acquisition unit or a first data acquisition unit. The preprocessing unit 12 may also be referred to as a learning-time preprocessing unit or a first preprocessing unit.

[0020] As shown in FIG. 1, the storage unit 90 may store the performance data DZ and the working day data DS used in the learning phase. Hereinafter, for example, the performance data DZ is simply abbreviated as DZ. The data acquisition unit 11 acquires DZ and DS as input data sets in the learning phase. Then, the data acquisition unit 11 divides the input data set into training data and verification data.

[0021] DZ0 in FIG. 2 is an example of the performance data as training data. DZ0 is a part of DZ. In the example of Embodiment 1, it is assumed that the unit of the required power amount recorded in the performance data is kWh. DS0 in FIG. 3 is an example of the working day data as training data. DS0 is a part of DS. In the example of Embodiment 1, it is assumed that the employment status recorded in the working day data has been previously converted into a qualitative variable (more specifically, a nominal scale) called "working day". In the example of FIG. 3, the value "1" of the working day indicates that the day (the day corresponding to the working day) is a working day. On the other hand, the value "0" of the working day indicates that the day is a holiday.

[0022] However, the types of employment status recorded in the working day data are not limited to the above example. For example, a type of employment status called "long vacation day" may be recorded in the working day data. A long vacation day means a holiday belonging to a predetermined long vacation (e.g., summer vacation period). In this case, for example, the value " -1" of the working day may indicate that the day is a long vacation day.

[0023] Before the learning unit 13 executes the machine learning algorithm, the preprocessing unit 12 generates preprocessed training data by performing a series of preprocessing on the training data. Next, the learning unit 13 generates a prediction model based on the preprocessed training data by executing the machine learning algorithm.

[0024] First, the preprocessing unit 12 generates new data by performing forward value interpolation on the working days recorded in DS0. DS1 in FIG. 4 is an example of the new data. DS1 may be referred to as the working day data after forward interpolation. Specifically, the preprocessing unit 12 converts DS0 into new data having the same time resolution (e.g., 1 minute) as DZ0 by forward interpolation. As shown in FIG. 4, as a result of the forward value interpolation, in DS1, the working days belonging to the same date have the same value at all time points.

[0025] Next, the preprocessing unit 12 generates new data by combining DZ0 and DS1 using the time point as a key. DZ1 in FIG. 4 is an example of the new data. For example, the preprocessing unit 12 generates DZ1 by adding the "working day" column in DS1 to DZ0. DZ1 may be referred to as the actual performance data with working day data.

[0026] Next, the preprocessing unit 12 derives the partial autocorrelation coefficient of the power demand amount in DZ1. In this specification, the autocorrelation coefficient is abbreviated as PAC. Specifically, the preprocessing unit 12 varies the lag variously and calculates the PAC for each lag. The lag represents the temporal delay amount set for the feature amount as time series data. The preprocessing unit 12 may further derive |PAC| which is the absolute value of PAC. The preprocessing unit 12 may further derive the autocorrelation coefficient of the power demand amount in DZ1. In this specification, the autocorrelation coefficient is abbreviated as AC. In Embodiment 1, the case where the preprocessing unit 12 derives PAC, |PAC|, and AC is exemplified.

[0027] The preprocessing unit 12 may set a lag as the product of the lag unit amount and the lag order. Then, the preprocessing unit 12 may calculate PAC, |PAC|, and AC for each determined lag. As an example, the lag unit amount may be a constant value preset by the user of the information processing apparatus 1. In the example of Embodiment 1, the case where the lag unit amount is 30 minutes is illustrated. However, the lag unit amount may be set by the preprocessing unit 12. For example, the preprocessing unit 12 may set the time resolution of DZ1 (e.g., 1 minute) as the lag unit amount.

[0028] The preprocessing unit 12 may generate a table showing the correspondence between the lag and PAC. The preprocessing unit 12 may also generate a table showing the correspondence between the lag, PAC, and |PAC|. The preprocessing unit 12 may also generate a table showing the correspondence between the lag and AC. Therefore, the preprocessing unit 12 may also generate a table showing the correspondence between the lag, PAC, |PAC|, and AC. LT in FIG. 6 shows an example of such a table.

[0029] In the example of FIG. 6, the lag order may be used as the key in LT. In the example of FIG. 6, the preprocessing unit 12 sets 1344 lags by changing the lag order from 1 to 1344. Therefore, the preprocessing unit 12 calculates 1344 PACs, |PAC|s, and ACs. The lag in the example of Embodiment 1 is an amount representing the time length per day.

[0030] As described above, since the lag unit amount is 30 minutes, lag order 1 corresponds to 30 minutes. Therefore, lag order 48 corresponds to 1 day (= 30 minutes × 48). And lag order 336 (= 48 × 7) corresponds to 7 days, that is, 1 week. And lag order 672 (= 336 × 2) corresponds to 14 days, that is, 2 weeks, lag order 1008 (= 336 × 3) corresponds to 21 days, that is, 3 weeks, and lag order 1344 (= 336 × 4) corresponds to 28 days, that is, 4 weeks.

[0031] In this specification, the element numbers of the data structure are generically represented by the integer i. In the example of FIG. 6, i may be a value equal to the lag order. Therefore, in the example of FIG. 6, the lag corresponding to i is denoted as lag[i]. Also, the PAC corresponding to i is denoted as PAC[i]. Accordingly, |PAC| corresponding to i is denoted as |PAC[i]|. Also, the AC corresponding to i is denoted as AC[i].

[0032] As is clear from the above description, lag[i] where i > 1 represents the i-th shortest lag. In this specification, unless otherwise inconsistent, the minimum value of the element number of the data structure is assumed to be 0 according to the notation of a general programming language. Although not shown in FIG. 6, as is clear to those skilled in the art, lag[0] = 0, PAC[0] = 1, and AC[0] = 1.

[0033] The inventors derived the relationship between the lag and the AC in a certain factory using the performance data obtained in that factory. Reference numeral 701 in FIG. 7 is a graph showing an example of the relationship between the lag and the AC in that factory. In addition, the inventors further derived the relationship between the lag and the PAC in that factory using the performance data. Reference numeral 702 in FIG. 7 is a graph showing an example of the relationship between the lag and the PAC in that factory. The horizontal axis in the graphs of reference numerals 701 and 702 both indicates the lag order. The vertical axis in the graph of reference numeral 701 indicates the AC, and the vertical axis in the graph of reference numeral 702 indicates the PAC. The horizontal axis in the graph of FIG. 7 can be read as the lag. The trend of the graph in FIG. 7 generally coincides with the example of FIG. 6.

[0034] As shown by reference numeral 701 in FIG. 7, the change in the AC corresponding to the lag is considered to correspond to the periodicity of the temporal variation of the demand power amount. Therefore, for example, the preprocessing unit 12 may select (extract) a special lag from among a plurality of set lags based on the AC. The special lag means, for example, a lag expected to represent some periodicity regarding the temporal variation of the demand power amount.

[0035] Also, as shown by reference numeral 702 in FIG. 7, the change in PAC corresponding to the lag is also considered to correspond to the periodicity of the temporal variation of the demand power amount. Therefore, for example, the preprocessing unit 12 may select a special lag from among the plurality of set lags based on the PAC.

[0036] In the example of reference numeral 701 in FIG. 7, generally for all lags, the absolute value of AC is relatively large. And the number of ACs having an absolute value close to 0 is relatively small. Further, the absolute value of the minimum value of AC is smaller than the absolute value of the maximum value of AC. On the other hand, in the example of reference numeral 702 in FIG. 7, generally for all lags, |PAC| is relatively small. And the number of PACs having an absolute value close to 0 is relatively large. Further, the absolute value of the minimum value of PAC is larger than the absolute value of the maximum value of AC. Therefore, the graph of PAC, unlike the graph of AC, has a sharp spike portion facing downward.

[0037] As described above, the graph of PAC can be significantly different from the graph of AC. However, in the example of FIG. 7, the lags at which the spike portions of PAC occur generally coincide with the lags at which the maximum values of AC occur. Therefore, it is expected that there is some relationship between PAC and AC.

[0038] Therefore, the preprocessing unit 12 may select a special lag from among the plurality of set lags based on PAC and AC. In Embodiment 1, the case where the preprocessing unit 12 selects a special lag based on PAC and AC will be exemplified.

[0039] As an example, the preprocessing unit 12 may select, as a first special lag candidate, a lag having a value equal to or greater than a time length threshold value among the plurality of lags and corresponding to |PAC| having a value equal to or greater than a partial autocorrelation threshold value.

[0040] Fluctuations in the required power within the facility are generally considered to have a cycle of approximately one day in the short term. Therefore, the time length threshold may be set to a value that is considered appropriate for grasping the short-term trend of fluctuations in the required power. The time length threshold may be a value that can be arbitrarily set by the user. In Embodiment 1, the case where the time length threshold is one day is illustrated.

[0041] The partial autocorrelation threshold may be set, for example, so that the spike portion of the PAC can be detected. In the example of reference numeral 702 in FIG. 7, |PAC| in the spike portion is generally 0.1 or more. From this, as an example, the partial autocorrelation threshold may be 0.05 to 0.15. The partial autocorrelation threshold may also be a value that can be arbitrarily set by the user. In Embodiment 1, the case where the partial autocorrelation threshold is 0.1 is illustrated.

[0042] As is obvious to those skilled in the art, when the lag is short, a PAC having a fairly large positive value is likely to occur. Therefore, by setting the time length threshold as described above, it is possible to avoid many short lags from being selected as the first special lag candidates. That is, the number of selected first special lag candidates can be reduced (see the rows with lag orders 1 to 3 in the example of FIG. 6). As a result, as can be understood from the following description, the quality of the prediction model can be improved.

[0043] As described above, the time length threshold of one day corresponds to i = 48. In the example of FIG. 6, in the range of i ≧ 48, |PAC

[48] | = 0.29 |PAC

[0336] | = 0.19 |PAC

[0672] | = 0.17 has a value equal to or greater than the partial autocorrelation threshold. Therefore, the preprocessing unit 12 lag

[48] = 1 lag

[0336] = 7 lag

[0672] = 14 is selected as the first special lag candidate.

[0044] Further, the preprocessing unit 12 may select, as the second special lag candidates, lags having values equal to or greater than the time length threshold among a plurality of lags and corresponding to the ACs having values equal to or greater than the autocorrelation threshold.

[0045] Generally, in the field of machine learning, when the correlation coefficient between two pieces of data is 0.7 or more, it can be considered that there is a strong positive correlation between the data. Therefore, as an example, the autocorrelation threshold may be 0.7 to 1. The autocorrelation threshold may also be a value that can be arbitrarily set by the user. In Embodiment 1, the case where the autocorrelation threshold is 0.7 is exemplified.

[0046] As is obvious to those skilled in the art, when the lag is short, an AC close to 1 is likely to occur. Therefore, by setting the time length threshold as described above, it is possible to avoid many short lags from being selected as the second special lag candidates. That is, the number of selected second special lag candidates can be reduced (see the rows of lag orders 1 to 3 in the example of FIG. 6).

[0047] In the example of FIG. 6, in the range of i ≧ 48, AC

[48] =0.78 AC

[0336] =0.73 AC

[0672] =0.70 AC

[1008] =0.72 have values equal to or greater than the autocorrelation threshold. Therefore, the preprocessing unit 12 lag

[48] =1 lag

[0336] =7 lag

[0672] =14 lag

[1008] =21 selects them as the second special lag candidates.

[0048] The preprocessing unit 12 may select a special lag based on the first special lag candidate and the second special lag candidate. As an example, the preprocessing unit 12 may select a lag common to the first special lag candidate and the second special lag candidate as the special lag. In the example of Embodiment 1, the second special lag candidate includes all of the first special lag candidates. Therefore, the preprocessing unit 12 selects the first special lag candidate as the special lag. That is, the preprocessing unit 12 selects the lags "7, 14, and 21" as the special lags.

[0049] Next, the preprocessing unit 12 may sort the selected plurality of special lags in ascending order. Then, the preprocessing unit 12 may generate an array Plag that includes the sorted special lags as elements. In the example of Embodiment 1, the preprocessing unit 12 generates Plag = [1, 7, 14]. Then, the preprocessing unit 12 selects elements that form an arithmetic progression (arithmetic progression elements) from among the elements of Plag. The method for selecting arithmetic progression elements will be described later.

[0050] In the example of Embodiment 1, the preprocessing unit 12 selects, as arithmetic progression elements, elements Plag[1:] = [7, 14] that are elements after the first element among the elements of Plag. These arithmetic progression elements form an arithmetic progression with the first term value of 7 and a common difference of 7.

[0051] Then, the preprocessing unit 12 selects, as remaining elements, elements of Plag excluding the arithmetic progression elements. The remaining elements may be referred to as non-arithmetic progression elements (elements that do not form an arithmetic progression). In the example of Embodiment 1, the processing unit 12 selects, as the remaining element, the element Plag[0] = [1] that is the 0th element among the elements of Plag.

[0052] The preprocessing unit 12 derives a new feature amount by setting the special lag selected as described above with respect to the required power amount. For example, the preprocessing unit 12 may derive at least one of a rolling feature and a lag feature as the new feature amount.

[0053] Based on the above Figure 7, it is considered that the special lag corresponding to the arithmetic sequence element is suitable for expressing the relatively long-term periodicity of the fluctuations in the required power amount. Therefore, the preprocessing unit 12 may select the special lag corresponding to the arithmetic sequence element as a rolling lag, which is a special lag for deriving the rolling feature amount. In the example of Embodiment 1, the preprocessing unit 12 selects the lags "7 and 14" as the rolling lags. In the example of Embodiment 1, the preprocessing unit 12 derives the rolling feature amount based on the rolling lags.

[0054] Then, the preprocessing unit 12 may select the special lag corresponding to the remaining elements as a non-rolling lag, which is a special lag for deriving a feature amount different from the rolling feature amount. In the example of Embodiment 1, the preprocessing unit 12 selects the lag "1" as the non-rolling lag. It is considered that the non-rolling lag is suitable for expressing the relatively short-term periodicity of the fluctuations in the required power amount. Therefore, in the example of Embodiment 1, the preprocessing unit 12 derives the lag feature amount based on the non-rolling lag.

[0055] In Embodiment 1, the case where the lag feature amount is derived prior to the derivation of the rolling feature amount is exemplified. However, as is obvious to those skilled in the art, the rolling feature amount may be derived prior to the derivation of the lag feature amount. Each process in Embodiment 1 may be executed in any order as long as there is no particular contradiction.

[0056] The preprocessing unit 12 derives a lag feature amount as a new feature amount corresponding to the required power amount by setting a non-rolling lag with respect to the required power amount. DZ2 in FIG. 8 shows an example of data generated by adding a lag feature amount to DZ1. In the example of Embodiment 1, since the non-rolling lag is 1, the preprocessing unit 12 derives the "required power amount one day before" as the lag feature amount. Therefore, in the example of FIG. 8, the "required power amount one day before" belonging to the row at the time point "2021 / 4 / 1 0:00" represents the "required power amount at 2021 / 3 / 31 0:00". Also, the "required power amount one day before" belonging to the row at the time point "2022 / 3 / 31 23:59" represents the "required power amount at 2022 / 3 / 30 23:59".

[0057] Next, the preprocessing unit 12 derives a rolling feature amount as a new feature amount corresponding to the required power amount by setting a non-rolling lag with respect to the required power amount. First, the preprocessing unit 12 acquires, as components (rolling feature amount components) of the rolling feature amount corresponding to a certain time point, the required power amounts at time points before the rolling lag with respect to that time point. In the example of Embodiment 1, since the rolling lags are 7 and 14, the preprocessing unit 12 acquires the "required power amount seven days before" and the "required power amount fourteen days before" as rolling feature amount components. The preprocessing unit 12 adds the rolling feature amount components to DZ2.

[0058] Next, the preprocessing unit 12 derives a statistical value of the rolling feature amount components as the rolling feature amount. In Embodiment 1, the case where the statistical value is the average value is exemplified. Therefore, the preprocessing unit 12 derives the average value of the "required power amount seven days before" and the "required power amount fourteen days before". The preprocessing unit 12 further adds the average value to DZ2 as the rolling feature amount.

[0059] DZ3 in FIG. 9 shows an example of data generated by adding rolling feature quantity components and rolling feature quantities to DZ2. In the example of FIG. 9, "demand power consumption 7 days ago" and "demand power consumption 14 days ago" belonging to the row at the time point "2021 / 4 / 1 0:00" represent "demand power consumption at 2021 / 3 / 25 0:00" and "demand power consumption at 2021 / 3 / 18 0:00", respectively. And "average (7, 14 days ago) demand power consumption" belonging to the row at the time point "2021 / 4 / 1 0:00" represents the average value of "demand power consumption 7 days ago" and "demand power consumption 14 days ago" belonging to the row at the same time point.

[0060] After the acquisition of the rolling feature quantity, the rolling feature quantity component is considered to be no longer necessary. Therefore, the preprocessing unit 12 may delete the rolling feature quantity component from DZ3 after the acquisition of the rolling feature quantity. DZ4 in FIG. 10 shows an example of data generated by deleting the rolling feature quantity component from DZ3.

[0061] Next, the preprocessing unit 12 derives a time equivalent quantity, which is a feature quantity corresponding to the time point, by converting the time point recorded in DZ4 into a predetermined type. Then, the preprocessing unit 12 adds the time equivalent quantity to DZ4. As an example, the preprocessing unit 12 may set the time equivalent quantity as an amount representing the elapsed time from the reference time point "0:00" to the target time point in units of hours (h). DZ5 in FIG. 11 shows an example of data generated by adding the time equivalent quantity to DZ4.

[0062] In the example of FIG. 11, the elapsed time of 1 minute from "0:00" corresponds to a time equivalent quantity of 1 / 60 (≈0.01667). In the example of Embodiment 1, the time equivalent quantity is derived using only hours (h) and minutes (m) without considering the year (y), month (m), and day (d) of the time point. However, as is obvious to those skilled in the art, an amount derived as a type considering at least one of the year, month, and day may also be derived as the time equivalent quantity.

[0063] However, as is obvious to those skilled in the art, the time-corresponding amount does not necessarily have to be derived. This is because the prediction model according to one aspect of the present disclosure may be set so as to be able to output a predicted value of the power demand amount by using, as explanatory variables, new feature amounts derived by setting special lags with respect to the power demand amount.

[0064] Next, the preprocessing unit 12 determines the day of the week corresponding to the date at the time recorded in DZ5 by referring to, for example, calendar data. Then, the preprocessing unit 12 adds the day of the week to DZ5. DZ6 in FIG. 12 is an example of data generated by adding the day of the week to DZ5. In the example of FIG. 12, the preprocessing unit 12 generates DZ6 by adding the day of the week, which has been converted into a qualitative variable (nominal scale), to DZ5. In the example of Embodiment 1, days of the week 0 to 6 represent Sunday to Saturday in this order. Therefore, the day of the week 4 in the example of FIG. 12 represents Thursday.

[0065] As shown in FIG. 12, in DZ6, the power demand amount and new feature amounts derived based on the power demand amount can be larger values than, for example, feature amounts (calendar feature amounts) indicating the day of the week and working days. Therefore, the preprocessing unit 12 may perform scaling on each feature amount included in DZ6. Thereby, the numerical value ranges of the respective feature amounts can be made to be about the same, so that it becomes possible to generate a prediction model having higher performance.

[0066] In Embodiment 1, a case where the preprocessing unit 12 performs standardization as scaling is exemplified. The preprocessing unit 12 derives the average value (μ) and standard deviation (σ) of a certain feature amount included in DZ6. Then, the preprocessing unit 12 standardizes the feature amount by using the derived μ and σ. In this specification, μ and σ used for standardization before generation of the learning model are referred to as standardization parameters. DZ6S in FIG. 13 is an example of data generated by performing standardization on each feature amount of DZ6. In the example of Embodiment 1, DZ6 is used as training data after preprocessing.

[0067] The learning unit 13 generates a prediction model by machine learning based on the time-series data of the required power amount after adding new feature amounts derived based on the required power amount. Therefore, as an example, the learning unit 13 may generate a prediction model based on the pre-processed training data (e.g., DZ6S) by executing a predetermined machine learning algorithm.

[0068] The prediction model according to one aspect of the present disclosure may be a learned model that can output a predicted value of the required power amount at the prediction target time point as a target variable. Therefore, the type of machine learning algorithm is not particularly limited as long as it can solve a regression task. Examples of machine learning algorithms include a neural network (NN), a support vector machine (SVM), and a decision tree (DT).

[0069] In the example of Embodiment 1, the learning unit 13 generates a prediction model that derives the target variable from the explanatory variables by using (i) the required power amount at an arbitrary prediction target time point included in DZ6S as the true value (correct answer data) of the target variable and (ii) each feature amount other than the required power amount before the prediction target time point as the explanatory variable.

[0070] After generating the prediction model, the learning unit 13 may incorporate the normalization parameters of each feature amount into the prediction model. The learning unit 13 may further incorporate a rolling lag and a non-rolling lag into the prediction model.

[0071] The evaluation unit 14 evaluates the prediction model generated by the learning unit 13. Specifically, the evaluation unit 14 evaluates the prediction performance of the prediction model. Prior to the evaluation by the evaluation unit 14, the preprocessing unit 12 generates preprocessed verification data by performing a series of preprocessing on the verification data in the same format as the above-described training data. Therefore, in the example of Embodiment 1, the preprocessed verification data has a data structure in the same format as DZ6S. For this reason, in the preprocessed verification data in Embodiment 1, each feature amount is standardized using the standardization parameters incorporated in the prediction model.

[0072] The evaluation unit 14 evaluates the prediction model using the preprocessed verification data. Specifically, the evaluation unit 14 inputs the preprocessed verification data into the prediction model to cause the prediction model to output a predicted value. Then, the evaluation unit 14 derives an index value indicating the prediction accuracy of the prediction model based on the predicted value and the true value of the required power amount shown in the preprocessed verification data.

[0073] The index value may be any value used for evaluating the prediction accuracy in a regression task in the field of machine learning and is not particularly limited. Examples of the index value include the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE). After deriving the index value, the evaluation unit 14 may store the prediction model in the storage unit 90.

[0074] Note that when the index value is less than a predetermined threshold, the model generation device 10 may cause the learning unit 13 to generate the prediction model again. In this case, prior to the regeneration of the prediction model, for example, the hyperparameters of the prediction model may be reset. In this way, in order to compensate for the performance of the prediction model used in the prediction phase, the generation of the prediction model may be repeated until an index value equal to or greater than the threshold is obtained. In this case, the evaluation unit 14 may store the prediction model in which an index value equal to or greater than the threshold is obtained for the first time in the storage unit 90.

[0075] (Example of method for selecting elements of arithmetic progression) FIG. 14 is a flowchart showing an example of the flow of processing for selecting arithmetic sequence elements. First, prior to S1, the preprocessing unit 12 obtains the number of elements NPlag of the above-described array Plag. In the example of Embodiment 1, since Plag = [1, 7, 14], NPlag = 3.

[0076] In S1, the preprocessing unit 12 reflects whether NPlag is 2 or more. If NPlag is 2 or more (YES in S1), the process proceeds to S2. On the other hand, if NPlag is 1 or less (NO in S1), the preprocessing unit 12 ends the processing of FIG. 14. In this case, the preprocessing unit 12 determines that Plag does not contain arithmetic sequence elements.

[0077] In S2, the preprocessing unit 12 initializes the element number i for pointing to the elements of Plag to 0. In S3, the preprocessing unit 12 initializes a flag flg indicating that Plag contains arithmetic sequence elements to True (true).

[0078] In S4, the preprocessing unit 12 determines whether i is less than NPlag - 2. If i is less than NPlag - 2 (YES in S4), the process proceeds to S5. On the other hand, if i is NPlag - 2 or more (NO in S4), the process proceeds to S8.

[0079] In S5, the preprocessing unit 12 substitutes Plag[i] into a0. a0 is the first term of the assumed arithmetic sequence. Thus, in S5, the preprocessing unit 12 sets a0 to a value equal to Plag[i]. In S6, the preprocessing unit 12 substitutes Plag[i + 1] - a0 into d. d is the common difference of the assumed arithmetic sequence. Thus, in S6, the preprocessing unit 12 sets d to a value equal to Plag[i + 1] - a0. Next, in S7, the preprocessing unit 12 substitutes i + 2 into the element number j for pointing to the elements of Plag. Then, the process proceeds to S11.

[0080] S8 to S10 are processes paired with S5 to S7 respectively. In S8, the preprocessing unit 12 substitutes 0 for a0. That is, in S8, the preprocessing unit 12 sets a0 to 0. In S9, the preprocessing unit 12 substitutes Plag[i] - a0 for d. Thus, in S9, the preprocessing unit 12 sets d to a value equal to Plag[i] - a0. Next, in S10, the preprocessing unit 12 substitutes i + 1 for j. Then, it proceeds to S11.

[0081] In S11, the preprocessing unit 12 substitutes Plag[j] - Plag[j - 1] for next_d. next_d is the next common difference in the assumed arithmetic progression. Thus, the preprocessing unit 12 sets next_d to a value equal to Plag[j] - Plag[j - 1].

[0082] In S12, the preprocessing unit 12 determines whether next_d is equal to d. If next_d is equal to d (YES in S12), in S13, the preprocessing unit 12 increments j by 1. Then, it proceeds to S15. On the other hand, if next_d is not equal to d (NO in S12), in S14, the preprocessing unit 12 updates flg to False (false). Then, it proceeds to S16.

[0083] In S15, the preprocessing unit 12 determines whether j is greater than or equal to NPlag. If j is greater than or equal to NPlag (YES in S15), it proceeds to S16. On the other hand, if j is less than NPlag (NO in S15), it returns to S11. Therefore, the processes of S11 to S13 and S15 are repeated until j becomes equal to NPlag.

[0084] In S16, the preprocessing unit 12 determines whether flg is False. If flg is False (YES in S16), in S17, the preprocessing unit 12 increments i by 1. Then, it proceeds to S19. On the other hand, if flg is not False (NO in S16), that is, if flg is True, it proceeds to S18.

[0085] In S18, the preprocessing unit 12 substitutes Plag[i:], which is the element of Plag after the i-th element, into the array ap. The array ap is an array for storing arithmetic sequence elements. After the completion of S18, the preprocessing unit 12 ends the process of FIG. 14. If flg is True at the end of the process (in other words, if the array ap at the end of the process is not an empty array), the preprocessing unit 12 determines that Plag contains arithmetic sequence elements. In this case, the preprocessing unit 12 may output ap as the selection result of the arithmetic sequence elements.

[0086] In S19, the preprocessing unit 12 determines whether i is greater than or equal to NPlag. If i is greater than or equal to NPlag (YES in S19), the process proceeds to S20. On the other hand, if i is less than NPlag (NO in S19), the process returns to S3. Therefore, the processes of S3 to S16, S17, and S19 are repeated until i becomes equal to NPlag.

[0087] In S20, the preprocessing unit 12 substitutes an empty array into the array ap. That is, the preprocessing unit 12 sets the array ap to an empty array. After the completion of S20, the preprocessing unit 12 ends the process of FIG. 14. If flg is False at the end of the process (in other words, if the array ap at the end of the process is an empty array), the preprocessing unit 12 determines that Plag does not contain arithmetic sequence elements. In this case, the preprocessing unit 12 may output ap as an empty array as the selection result of the arithmetic sequence elements.

[0088] In the example of Embodiment 1, after i = 0 is set in S2, in S5, a0 = Plag[0] = 1 is set. Therefore, in S6, d = Plag[1] - a0 = 7 - 1 = 6 is set. And in S7, j = 2 is set. For this reason, in S11, next_d = Plag[2] - Plag[1] = 14 - 7 = 7 is set. In this case, since d is not equal to next_d, flg is set to False in S14. After that, in S17, i is incremented by 1.

[0089] In S3, flg is updated to True. In the example of Embodiment 1, NPlag = 3. Therefore, when i = 1, in S4, the condition i < NPlag - 2 is not satisfied, so in S8, a0 is set to 0. Next, in S9, d = Plag[1] - a0 = 7 - 0 = 7 is set. Then, in S10, j is set to 2. For this reason, in S11, next_d = Plag[2] - Plag[1] = 14 - 7 = 7 is set. Therefore, since d is equal to next_d and flg is True, in S18, Plag[1:] = [7, 14] is set as ap.

[0090] (Prediction Phase) Subsequently, referring to FIG. 1 again, an example of the prediction phase following the learning phase will be described. The prediction device 20 may include a data acquisition unit 21, a preprocessing unit 22, a prediction calculation unit 23, and a prediction result output unit 24. The data acquisition unit 21 may also be referred to as a prediction-time data acquisition unit or a second data acquisition unit. The preprocessing unit 22 may also be referred to as a prediction-time preprocessing unit or a second preprocessing unit.

[0091] As shown in FIG. 1, the storage unit 90 may store the performance data DZP and the working day data DSP used in the prediction phase. In the prediction phase, real-time prediction of the required power amount may be performed. For this reason, the performance data DZP may be updated in real time. The data acquisition unit 21 acquires DZP and DSP as input data in the prediction phase. Assume that DZP has the same data structure as DZ. Also, assume that DSP has the same data structure as DS. In addition, the data acquisition unit 21 acquires the prediction model stored in the storage unit 90 by the model generation device 10 in the learning phase.

[0092] The preprocessing unit 22 generates preprocessed input data by performing a series of preprocessing in the same manner as the above-described learning phase on the input data set in the prediction phase. An example of the processing of the preprocessing unit 22 will be described below.

[0093] First, the preprocessing unit 22 generates pre-interpolation employment day data (for convenience, referred to as DSP1) in the prediction phase by performing previous value interpolation on the employment days in the DSP. As a result of the previous interpolation, DSP1 has the same time resolution (e.g., 1 minute) as DZP.

[0094] Next, the preprocessing unit 22 generates actual performance data with employment day data (for convenience, referred to as DZP1) in the prediction phase by combining DZP and DSP1 using the time point as a key.

[0095] Next, the preprocessing unit 22 acquires the special lags determined in the learning phase. For example, the preprocessing unit 22 may read out the rolling lag and non-rolling lag determined in the learning phase from the prediction model. The preprocessing unit 22 derives new feature quantities in the prediction phase by setting the special lags determined in the learning phase for the power demand amount in the prediction phase.

[0096] The preprocessing unit 22 may derive a lag feature quantity as a new feature quantity in the prediction phase by setting a non-rolling lag for the power demand amount in the prediction phase. Then, the preprocessing unit 22 may generate new data (for convenience, referred to as DZP2) by adding the lag feature quantity to DZP1. In the example of Embodiment 1, since the non-rolling lag is 1, the preprocessing unit 22 acquires "the power demand amount one day before" as the lag feature quantity in the prediction phase.

[0097] Next, the preprocessing unit 22 may derive a rolling feature amount as a new feature amount in the prediction phase by setting a rolling lag for the required power amount in the prediction phase. First, the preprocessing unit 22 acquires a rolling feature amount component in the prediction phase according to the rolling lag. In the example of Embodiment 1, since the rolling lags are 7, 14, and 21, the preprocessing unit 22 acquires the "required power amount 7 days ago", "required power amount 14 days ago", and "required power amount 21 days ago" in the prediction phase as rolling feature amount components in the prediction phase. Then, the preprocessing unit 22 adds the rolling feature amount component to DZP2.

[0098] Next, the preprocessing unit 22 derives the average value of the "required power amount 7 days ago", the "required power amount 14 days ago", and the "required power amount 21 days ago" in the prediction phase, and acquires the average value as a rolling feature amount in the prediction phase. Then, the preprocessing unit 22 further adds the rolling feature amount to DZP2.

[0099] As described above, the preprocessing unit 22 may generate new data (referred to as DZP3 for convenience) by adding the rolling feature amount component and the rolling feature amount in the prediction phase to DZP2. After acquiring the rolling feature amount, the preprocessing unit 22 may generate new data (referred to as DZP4 for convenience) by deleting the rolling feature amount component from DZP3.

[0100] Next, the preprocessing unit 22 converts the time point recorded in DZP4 into an equivalent time point in the prediction phase. Then, the preprocessing unit 22 generates new data (referred to as DZP5 for convenience) by adding the equivalent time point to DZ4.

[0101] Next, the preprocessing unit 22 generates new data (referred to as DZP6 for convenience) by adding the day of the week (the day of the week converted into a qualitative variable) corresponding to the date of the time point recorded in DZP5 to DZ5.

[0102] Next, the preprocessing unit 22 reads out the normalization parameters determined in the learning phase from the prediction model. Then, the preprocessing unit 22 normalizes each feature quantity of DZP6 using the normalization parameters. DZP6S in FIG. 15 is an example of data generated by performing normalization on each feature quantity of DZP6. DZP6S is used as the preprocessed input data in the prediction phase. In FIG. 15, the illustration of the time point as the key is omitted. That is, in FIG. 15, only the explanatory variables used for predicting the required power consumption in the prediction phase are shown.

[0103] As an example, in the prediction phase, the prediction of the required power consumption may be executed every 30 minutes. Also, in the prediction phase, for example, the predicted value of the required power consumption over a future period exceeding 24 hours (e.g., up to 36 hours after the current time) may be derived. From this, in the prediction phase, for example, the values of some cells of "the required power consumption one day ago" may be unknown. In the example of FIG. 15, cells having unknown values (i.e., blank values) are shown by the notation NULL. When there are blank values in DZP6S, the preprocessing unit 22 may interpolate the blank values using a predetermined interpolation method.

[0104] As an example, the preprocessing unit 22 may interpolate the blank values based on the known values of the required power consumption one day ago using the prediction model. For example, the preprocessing unit 22 may input the known value of the required power consumption one day ago corresponding to a certain blank value into the prediction model, causing the prediction model to output a predicted value corresponding to the input. Then, the preprocessing unit 22 may set the predicted value as the interpolated value of the blank value.

[0105] The prediction calculation unit 23 causes the prediction model to output a predicted value of the power demand based on the time series data of the power demand after a new feature amount derived based on the power demand in the prediction phase is added. Therefore, the prediction calculation unit 23 may cause the prediction model to output a predicted value by inputting the pre-processed input data (e.g., DZP6S) to the prediction model. Specifically, the prediction calculation unit 23 supplies each feature amount in the pre-processed input data to the prediction model as an explanatory variable. The prediction model outputs a predicted value corresponding to the explanatory variable as an objective variable.

[0106] The prediction result output unit 24 generates prediction result data based on the predicted value of the power demand output by the prediction model. For example, the prediction result output unit 24 may generate data associating each predicted value with each time point as the prediction result data. The prediction result output unit 24 outputs the generated prediction result data.

[0107] Also, the prediction result output unit 24 may convert the predicted value of the power demand (unit: kWh) to another unit. As an example, the prediction result output unit 24 may convert the predicted value of the power demand to the predicted value of the power demand (unit: kW). In this case, the prediction result output unit 24 may generate data associating each converted predicted value with each time point as the prediction result data.

[0108] (Effect of the information processing apparatus 1) According to the model generation apparatus 10, in the learning phase, a special lag can be selected based on the PAC (partial autocorrelation coefficient of the power demand). Then, by setting the special lag for the power demand, a new feature amount can be derived. Thus, according to the model generation apparatus 10, the new feature amount can be derived based on the PAC from the feature amounts expected to represent the periodicity of the fluctuations in the power demand (e.g., the above-described rolling feature amount and lag feature amount). Therefore, it becomes possible to obtain the new feature amount as an explanatory variable expected to be beneficial for predicting the power demand.

[0109] According to the model generation device 10, a prediction model can be generated based on time series data in which new feature quantities are added to the required power consumption (for example, refer to DZ4 in FIG. 9 described above). Therefore, since a prediction model can be generated using the new feature quantity as an explanatory variable, the prediction accuracy of the prediction model can be improved. That is, a high-quality prediction model can be obtained.

[0110] The fluctuation trend of the required power consumption in a facility may vary depending on, for example, the business type of the facility. Therefore, prior to generating a prediction model, it is not always easy for the user to manually set an appropriate lag according to the facility. Also, if the lag set by the user is inappropriate, the quality of the prediction model may deteriorate.

[0111] However, as described above, according to the model generation device 10, a special lag that is expected to be an appropriate lag for grasping the periodicity of the fluctuation of the required power consumption can be selected based on PAC. Therefore, a prediction model can be generated without requiring manual setting of the lag by the user. Thus, while enhancing the convenience for the user, a high-quality prediction model according to the business type of the facility can be obtained.

[0112] Furthermore, according to the model generation device 10, a prediction model can also be generated by using the employment status in the facility as an additional explanatory variable. By using the employment status as an additional explanatory variable, a prediction model that more specifically reflects the business type of the facility can be generated. Therefore, it becomes possible to further improve the prediction accuracy of the prediction model.

[0113] By the way, in Patent Document 1, an idea of using a plurality of NNs separately to improve the prediction accuracy of the heat load is shown. Specifically, in Patent Document 1, it is disclosed that three individual NNs, namely, an NN for weekdays, an NN for Saturdays, and an NN for Sundays, are generated in advance, and each NN is used separately according to the day of the week of the prediction date.

[0114] However, depending on the type of facility, at least one of Saturday and Sunday may be set as a working day. Alternatively, depending on the type of facility, a weekday (e.g., Monday) of a specific day of the week may be set as a holiday. Alternatively, depending on the type of facility, all of weekdays, Saturday, and Sunday may be applicable as working days.

[0115] Therefore, in the proper use of each NN shown in Patent Document 1, it is not possible to flexibly cope with the various types of facilities described above. From this, when the proper use of each NN shown in Patent Document 1 is applied to the demand power prediction, it is not always possible to achieve high prediction accuracy.

[0116] On the other hand, according to the model generation device 10, as described above, it is possible to obtain a high-quality prediction model according to the type of facility. Therefore, according to the model generation device 10, unlike the technology of Patent Document 1, it is possible to generate a single prediction model capable of achieving high prediction accuracy. Thus, according to the model generation device 10, a prediction model with higher versatility can be obtained compared to the conventional one.

[0117] And according to the prediction device 20, in the prediction phase, it is possible to derive a predicted value of the demand power using the prediction model generated by the model generation device 10 in the learning phase. As described above, according to one aspect of the present invention, it is possible to improve the prediction accuracy of the demand power in a facility compared to the conventional one.

[0118] (Supplementary Note 1) Depending on the type of facility, a case (referred to as the first case for convenience) where many spike portions occur in the PAC can also be considered. In the first case, when a special lag is selected based only on the PAC, many special lags can be selected. In this case, many new explanatory variables (e.g., many rolling feature amounts and lag feature amounts) are derived. As is obvious to those skilled in the art, when the number of explanatory variables is excessive, the quality of the prediction model may deteriorate. Also, an excessive number of explanatory variables causes an increase in the time required for the execution of the machine learning algorithm.

[0119] Therefore, as described above, the model generation device 10 may select a special lag based on PAC and AC (the autocorrelation coefficient of the required power consumption) in the learning phase. Thereby, compared with the case where the special lag is selected based only on PAC in the first case, the number of special lags selected in the first case can be reduced.

[0120] Therefore, by selecting a special lag based on PAC and AC, the number of newly derived explanatory variables can be reduced compared to the case where the special lag is selected based only on PAC. As a result, a high-quality prediction model can be generated even in the first case. In addition, the time required to execute the machine learning algorithm can also be reduced.

[0121] (Supplementary Note 2) Depending on the type of facility, there may also be a case (hereinafter referred to as the second case for convenience) where there are many lags corresponding to AC having a value equal to or greater than the autocorrelation threshold in the vicinity of the lag corresponding to the maximum value of AC. Therefore, for example, the preprocessing unit 12 may first select, from among a plurality of lags, the lag corresponding to the maximum value of AC in the data series indicating the relationship between the lag and AC. Next, the preprocessing unit 12 may select, as a second special lag candidate, a lag having a value equal to or greater than the time length threshold among the lags corresponding to the maximum value of AC and corresponding to AC having a value equal to or greater than the autocorrelation threshold.

[0122] According to the above-described method for selecting the second special lag candidate, the number of second special lags selected can be reduced compared to the method for selecting the second special lag candidate exemplified in Embodiment 1. As a result, for example, the number of special lags selected in the second case can be reduced. Therefore, a high-quality prediction model can be generated even in the second case. In addition, the time required to execute the machine learning algorithm can also be reduced.

[0123] [Embodiment 2] In Embodiment 1, a method of selecting a common lag between the first special lag candidate and the second special lag candidate as the special lag was exemplified. In Embodiment 2, a method of selecting a special lag different from that in Embodiment 1 will be described.

[0124] In Embodiment 2, the preprocessing unit 12 derives the sum SUM of |PAC| and AC corresponding to each of a plurality of lags. That is, the preprocessing unit 12 SUM = |PAC| + AC to derive.

[0125] The preprocessing unit 12 may select a special lag based on SUM. As an example, the preprocessing unit 12 may select, as a special lag, a lag among the plurality of lags that has a value equal to or greater than a time length threshold value and corresponds to a SUM that has a value equal to or greater than a sum threshold value.

[0126] The sum threshold value may be set according to the partial autocorrelation threshold value and the autocorrelation threshold value described in Embodiment 1. As an example, the sum threshold value may be set to a value equal to the sum of the partial autocorrelation threshold value and the autocorrelation threshold value. For this reason, for example, the sum threshold value may be 0.75 to 1.15. As is clear from the description of Embodiment 1, the sum threshold value may also be a value that can be arbitrarily set by the user. In Embodiment 2, the case where the sum threshold value is 0.8 is illustrated.

[0127] In this specification, the SUM corresponding to i described in Embodiment 1 is denoted as SUM[i]. In the example of FIG. 6 described above, in the range where i ≥ 48, SUM

[48] = 1.07 SUM

[0336] = 0.92 SUM

[0672] = 0.87 SUM

[1008] = 0.81 has a value equal to or greater than the sum threshold value. Therefore, in Embodiment 2, the preprocessing unit 12 lag

[48] = 1 lag

[0336] = 7 lag

[0672] = 14 lag

[1008] = 21 as special lags.

[0128] 〔Modification Example〕 As is obvious to those skilled in the art, the explanatory variables in the prediction model are not limited to the examples of the above-described embodiments. Any feature quantity that can be considered to affect the required power consumption in the facility may be used as an explanatory variable. For example, depending on the type of facility, meteorological conditions such as temperature, weather, and solar radiation affect the required power consumption. Therefore, when there is a correlation between the meteorological conditions and the required power consumption, a feature quantity indicating the meteorological conditions (meteorological feature quantity) can also be used as an additional explanatory variable.

[0129] The meteorological feature quantity may be an actual value obtained as a result of measurement in the facility, or may be a predicted value provided by a meteorological forecast service provider. Therefore, for example, in the learning phase, a prediction model may be generated using the actual value of the meteorological feature quantity. Next, in the prediction phase, the required power consumption may be predicted by the prediction model using the predicted value of the meteorological feature quantity. In this case, in the prediction phase, the predicted value of the meteorological feature quantity may be corrected using the actual value of the meteorological feature quantity in the learning phase.

[0130] 〔Example of Realization by Software〕 The functions of the information processing apparatus 1 (hereinafter referred to as the "apparatus") can be realized by a program for causing a computer to function as the apparatus, and by a program for causing a computer to function as each control block of the apparatus (particularly each part included in the control apparatus 9).

[0131] In this case, the above-described apparatus includes, as hardware for executing the above-described program, a computer having at least one control apparatus (for example, a processor) and at least one storage apparatus (for example, a memory). By executing the above-described program by this control apparatus and storage apparatus, each function described in the above-described embodiments is realized.

[0132] The above program may be recorded on one or more computer-readable recording media, rather than temporarily. This recording medium may or may not be provided in the above device. In the latter case, the above program may be supplied to the above device via any wired or wireless transmission medium.

[0133] Also, part or all of the functions of each of the above control blocks can also be realized by a logic circuit. For example, an integrated circuit in which a logic circuit functioning as each of the above control blocks is formed is also included in the scope of one aspect of the present invention. In addition to this, for example, it is also possible to realize the functions of each of the above control blocks by a quantum computer.

[0134] As is clear from the description of each of the above embodiments, each process in each of the above embodiments can be executed by AI (Artificial Intelligence). In this case, the AI may operate in the above control device, or may operate in another device (for example, an edge computer or a cloud server, etc.).

[0135] 〔Summary〕 The model generation device according to Aspect 1 of the present invention is a model generation device that generates a prediction model for predicting the required power amount in a facility. The model generation device sets a plurality of lags with respect to the actual value for time-series data indicating the time transition of the actual value of the required power amount, thereby deriving the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags, and based on the partial autocorrelation coefficient, selects a special lag from among the plurality of lags, and by setting the special lag with respect to the actual value, a new feature amount is derived, and based on the time-series data with the new feature amount added, the prediction model is generated by machine learning.

[0136] The model generation device according to Aspect 2 of the present invention may further derive the autocorrelation coefficient of the actual value corresponding to each of the plurality of lags in Aspect 1, and may select the special lag based on the partial autocorrelation coefficient and the autocorrelation coefficient.

[0137] In the model generation device according to Embodiment 3 of the present invention, in the above Embodiment 2, the absolute value of the partial autocorrelation coefficient corresponding to each of the plurality of the lags may be further derived. Among the plurality of the lags, a lag having a value equal to or greater than the time length threshold value and having an absolute value equal to or greater than the partial autocorrelation threshold value may be selected as a first special lag candidate. Among the plurality of the lags, a lag having a value equal to or greater than the time length threshold value and having an autocorrelation coefficient having a value equal to or greater than the autocorrelation threshold value may be selected as a second special lag candidate. Among the plurality of the lags, a lag common to the first special lag candidate and the second special lag candidate may be selected as the special lag.

[0138] In the model generation device according to Embodiment 4 of the present invention, in the above Embodiment 3, the partial autocorrelation threshold value may be a value of 0.05 or more and 0.15 or less, and the autocorrelation threshold value may be a value of 0.7 or more and 1 or less.

[0139] In the model generation device according to Embodiment 5 of the present invention, in the above Embodiment 2, the sum of the absolute value of the partial autocorrelation coefficient corresponding to each of the plurality of the lags and the autocorrelation coefficient may be further derived. Among the plurality of the lags, a lag having a value equal to or greater than the time length threshold value and having a sum equal to or greater than the sum threshold value may be selected as the special lag.

[0140] In the model generation device according to Embodiment 6 of the present invention, in the above Embodiment 5, the sum threshold value may be a value of 0.75 or more and 1.15 or less.

[0141] In the model generation device according to Embodiment 7 of the present invention, in any one of the above Embodiments 1 to 5, in an array in which a plurality of the special lags are sorted in ascending order, a plurality of the special lags forming an arithmetic progression may be selected as rolling lags, and by setting the rolling lags with respect to the actual values, rolling feature amounts may be derived as the new feature amounts.

[0142] In the model generation device according to aspect 8 of the present invention, in the above aspect 7, the special lag excluding the rolling lag in the above array may be selected as a non-rolling lag, and by setting the non-rolling lag with respect to the above actual value, a lag feature amount may be derived as the above new feature amount.

[0143] In the model generation device according to aspect 9 of the present invention, in any one of the above aspects 1 to 8, the time series data with the above new feature amount added may further include data indicating the daily employment status in the above facility.

[0144] The prediction device according to aspect 10 of the present invention is a prediction device that derives a predicted value of the above demand power consumption by using, in a prediction phase, a prediction model for predicting the demand power consumption in a facility, which was generated in advance in a learning phase. In the above learning phase, for the time series data in the above learning phase indicating the time transition of the actual value of the above demand power consumption, by setting a plurality of lags with respect to the above actual value, the partial autocorrelation coefficient of the above actual value corresponding to each of the plurality of the above lags is derived. Based on the above partial autocorrelation coefficient, a special lag is selected from among the plurality of the above lags, and by setting the special lag with respect to the above actual value, a new feature amount is derived. Based on the above time series data with the above new feature amount added, the above prediction model is generated by machine learning. The above prediction device sets the above special lag with respect to the above actual value in the above prediction phase for the time series data in the above prediction phase indicating the time transition of the above actual value, to derive a new feature amount in the above prediction phase, and based on the above time series data in the above prediction phase with the above new feature amount added in the above prediction phase, causes the above predicted value to be output to the above prediction model.

[0145] The model generation method according to Aspect 11 of the present invention is a model generation method for generating a prediction model for predicting the required power amount in a facility, and for time series data showing the temporal transition of the actual value of the required power amount, by setting a plurality of lags with respect to the actual value, a step of deriving the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags; a step of selecting a special lag from among the plurality of lags based on the partial autocorrelation coefficient; a step of deriving a new feature amount by setting the special lag with respect to the actual value; and a step of generating the prediction model by machine learning based on the time series data to which the new feature amount is added.

[0146] The prediction method according to Aspect 12 of the present invention is a prediction method for deriving a predicted value of the required power amount by using, in a prediction phase, a prediction model for predicting the required power amount in a facility, which was generated in advance in a learning phase, wherein in the learning phase, for the time series data in the learning phase showing the temporal transition of the actual value of the required power amount, by setting a plurality of lags with respect to the actual value, the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags has been derived, based on the partial autocorrelation coefficient, a special lag has been selected from among the plurality of lags, a new feature amount has been derived by setting the special lag with respect to the actual value, and the prediction model has been generated by machine learning based on the time series data to which the new feature amount is added, and the prediction method includes a step of deriving a new feature amount in the prediction phase by setting the special lag with respect to the actual value in the prediction phase for the time series data in the prediction phase showing the temporal transition of the actual value; and a step of causing the prediction model to output the predicted value based on the time series data in the prediction phase to which the new feature amount in the prediction phase is added.

[0147] 〔Supplementary Notes〕 One aspect of the present invention is not limited to each of the above-described embodiments, and various modifications are possible within the scope shown in the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of one aspect of the present invention.

Explanation of Reference Numerals

[0148] 1 Information processing apparatus 10 Model generation apparatus 11 Data acquisition unit (data acquisition unit of model generation apparatus) 12 Preprocessing unit (preprocessing unit of model generation apparatus) 13 Learning unit 14 Evaluation unit 20 Prediction apparatus 21 Data acquisition unit (data acquisition unit of prediction apparatus) 22 Preprocessing unit (preprocessing unit of prediction apparatus) 23 Prediction calculation unit 24 Prediction result output unit

Claims

1. A model generation device for generating a prediction model for predicting the required power consumption in a facility, The model generation device, For time series data showing the time transition of the actual value of the required power consumption, by setting a plurality of lags with respect to the actual value, the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags is derived, Based on the partial autocorrelation coefficient, a special lag is selected from among the plurality of lags, By setting the special lag with respect to the actual value, a new feature amount is derived, A model generation device that generates the prediction model by machine learning based on the time series data with the new feature amount added.

2. The model generation device, Further derives the autocorrelation coefficient of the actual value corresponding to each of the plurality of lags, The model generation device according to claim 1, wherein the special lag is selected based on the partial autocorrelation coefficient and the autocorrelation coefficient.

3. The model generation device, Further derives the absolute value of the partial autocorrelation coefficient corresponding to each of the plurality of lags, Among the plurality of lags, a lag having a value equal to or greater than the time length threshold value and corresponding to an absolute value having a value equal to or greater than the partial autocorrelation threshold value is selected as a first special lag candidate, Among the plurality of lags, a lag having a value equal to or greater than the time length threshold value and corresponding to an autocorrelation coefficient having a value equal to or greater than the autocorrelation threshold value is selected as a second special lag candidate, The model generation device according to claim 2, wherein a lag common to the first special lag candidate and the second special lag candidate among the plurality of lags is selected as the special lag.

4. The partial autocorrelation threshold value is a value of 0.05 or more and 0.15 or less, The model generation device according to claim 3, wherein the autocorrelation threshold value is a value of 0.7 or more and 1 or less.

5. The model generation device is further deriving the sum of the absolute value of the partial autocorrelation coefficient corresponding to each of the plurality of the lags and the autocorrelation coefficient, selecting, as the special lag, a lag having a value equal to or greater than the time length threshold value among the plurality of the lags and corresponding to the sum having a value equal to or greater than the sum threshold value, the model generation device according to claim 2.

6. The model generation device according to claim 5, wherein the sum threshold value is a value of 0.75 or more and 1.15 or less.

7. The model generation device is selecting, as the rolling lag, a plurality of the special lags forming an arithmetic progression in an array in which the plurality of the special lags are sorted in ascending order, deriving a rolling feature amount as the new feature amount by setting the rolling lag with respect to the actual value, the model generation device according to claim 1.

8. The model generation device is selecting, as the non-rolling lag, the special lags excluding the rolling lag in the array, deriving a lag feature amount as the new feature amount by setting the non-rolling lag with respect to the actual value, the model generation device according to claim 7.

9. The model generation device according to claim 1, wherein the time series data with the new feature amount added further includes data indicating the daily employment status in the facility.

10. A prediction device that derives a predicted value of the required power amount by using, in a prediction phase, a prediction model for predicting the required power amount in a facility, which is generated in advance in a learning phase, in the learning phase, For the time series data in the learning phase showing the time transition of the actual value of the required power amount, by setting a plurality of lags with respect to the actual value, the partial autocorrelation coefficients of the actual value corresponding to each of the plurality of lags are derived. Based on the partial autocorrelation coefficients, a special lag is selected from among the plurality of lags. By setting the special lag with respect to the actual value, a new feature amount is derived. Based on the time series data with the new feature amount added, a prediction model is generated by machine learning. The prediction device For the time series data in the prediction phase showing the time transition of the actual value, by setting the special lag with respect to the actual value in the prediction phase, a new feature amount in the prediction phase is derived. A prediction device that causes the prediction model to output a predicted value based on the time series data in the prediction phase with the new feature amount added in the prediction phase.

11. A model generation method for generating a prediction model for predicting the required power amount in a facility, which is generated by a model generation device, A step in which the model generation device derives partial autocorrelation coefficients of the actual value corresponding to each of a plurality of lags by setting a plurality of lags with respect to the actual value for time series data showing the time transition of the actual value of the required power amount; A step in which the model generation device selects a special lag from among the plurality of lags based on the partial autocorrelation coefficients; A step in which the model generation device derives a new feature amount by setting the special lag with respect to the actual value; A step in which the model generation device generates the prediction model by machine learning based on the time series data with the new feature amount added, the model generation method comprising the steps.

12. A prediction method in which a prediction device derives a predicted value of the required power consumption by using, in a prediction phase, a prediction model for predicting the required power consumption in a facility, which has been generated in advance by a model generation device in a learning phase, In the learning phase, The model generation device sets a plurality of lags with respect to the actual value for time-series data in the learning phase showing the time transition of the actual value of the required power consumption, and thus the partial autocorrelation coefficient of the actual value corresponding to each of the plurality of lags is derived, The model generation device selects a special lag from among the plurality of lags based on the partial autocorrelation coefficient, The model generation device sets the special lag with respect to the actual value, thereby deriving a new feature amount, The model generation device generates the prediction model by machine learning based on the time-series data with the new feature amount added, The prediction method includes: A step in which the prediction device sets the special lag with respect to the actual value in the prediction phase for time-series data in the prediction phase showing the time transition of the actual value, thereby deriving a new feature amount in the prediction phase; A step in which the prediction device causes the prediction model to output the predicted value based on the time-series data in the prediction phase with the new feature amount added in the prediction phase. The prediction method includes these steps.

Images