JP6746946B2 - Power demand forecasting apparatus and power demand forecasting method - Google Patents

Description

The present invention relates to a power demand prediction device and a power demand prediction method.

Patent Document 1 describes a method of predicting a change in power demand by identifying a past demand curve that is the closest to the demand curve of the day. Specifically, (a) the power demand for the current day within 2 hours until the current time is obtained at 5-minute intervals, and (b) the power demand database is referred to, and the power demand for 2 hours in any past time zone The amount is calculated at 5-minute intervals, (c) the deviation between the electric power demand on the day and the past electric power demand calculated in (b) is obtained, (d) the sum Sd of the absolute values of the deviations is obtained, and (e) It is described that the demand curve in the time zone in which Sd becomes the minimum is specified as the one that most closely approximates the demand curve of the day.

Patent Document 2 describes a power demand prediction support method for predicting daily power demand at a power company or the like. Specifically, it stores and manages past load data in a database, accepts input of load data at the present time, accepts the search target period and designation of a predetermined time period before the present time, and loads the load input in the specified time period. Compare the data with the past load data in the same time zone as the specified time zone in the search target period, extract multiple past load data similar to the fluctuation of the load data input in the specified time zone, and reduce the past short time Minute load data and a plurality of load curves for a predetermined number of past predetermined time periods are displayed and output, and user input to specify a specific curve from the plurality of past load change curves being output is displayed. It is described that only the designated fluctuation curve and the fluctuation curve of the load in the past short time are hidden and displayed.

In Patent Document 3, a prediction device receives an input of a prediction period of power prediction data and a classification period located at a time before the prediction period, and classifies the power demand data of the classification period on the prediction target day on the target classification. Data for the forecast period, the power demand data for the classification period past the forecast target date is read as data for the past category, and the power demand data for the forecast period past the forecast target date is read as the past forecast period data. Target data and past classification data that have been subjected to normalization calculation by using a predetermined average value and a predetermined standard deviation as predetermined values for the respective power demand data Based on the power demand data of, the data is classified into one of multiple classes, and the past classification data of the same class as the target classification data is extracted from the classified past classification data, and the extracted past classification It is described that the statistical value of the past forecast period data corresponding to the data for calculation is calculated as the target forecast period data which is the power forecast data in the forecast period on the day of the forecast target day, and the calculated target forecast period data is output. ..

JP, 2005-224002, A JP, 2006-178893, A JP, 2014-85914, A

In order to carry out stable and economical power supply and demand operation, it is necessary to formulate a power supply and demand plan based on accurate prediction of power demand. Further, in the recent electric power system reform, the same planned amount is required at the same time, and more accurate demand forecast is required.

On the other hand, the fluctuation factors of the power demand curve are diversifying, such as the change in daytime demand due to the large-scale introduction of renewable energy, especially solar power generation, and the change in late-night demand due to the increase in energy-saving equipment such as heat pump water heaters. It is becoming difficult to predict accurately.

When forecasting power demand, past time series data are used in various forecasting methods. However, it takes a great deal of time and effort to search for time series data close to the power demand on the predicted date from a huge amount of past time series data by, for example, a human system. In addition, for example, if an attempt is made to automatically search for time series data that is close to the power demand on the prediction date, time series data having different characteristics may be searched, and the prediction accuracy may drop significantly.

The present invention has been made in view of such a background, and an object of the present invention is to provide a power demand prediction device and a power demand prediction method capable of accurately predicting power demand.

One of the inventions for achieving the above object is a power demand prediction apparatus, which is a time series data storage unit that stores time series data that is an actual value of power demand, and a predetermined period of a prediction date. , Comparing the predicted date/time series data, which is the time series data in the predetermined period, with the past date/time series data, which is the time series data on a day before the predicted date, and the past in which the magnitude relationship is not reversed in the predetermined period the time series data, and the predetermined period said in preference to the past time-series data Increase Decrease relationship is reversed Te smell, selects as the time series data used for predicting the power demand, the time-series data selecting unit, selecting And a power demand prediction unit that predicts the power demand on the predicted day based on the time series data.

According to the present invention, it is possible to select time-series data having a tendency similar to that of the prediction date as time-series data used for power demand prediction, and to exclude it only by the determination based on the difference (distance). It is possible to exclude time series data with different characteristics that cannot be obtained. Therefore, it is possible to improve the prediction accuracy of the power demand, and eventually realize economical system operation and reduce the imbalance (difference between the power consumption and the generated power).

Another one of the present invention is the power demand prediction apparatus, wherein the time series data selection unit divides the predetermined period into a first half period and a second half period, and averages the forecast date/time series data in the first half period. The forecast first half average, the forecast date second half average that is the average of the forecast date/time series data in the second half period, the past day first half average that is the average of the past date/time series data in the first half period, and the past date/time in the second half period The average of the series data, the latter half of the past day, is obtained, and a positive value is taken if the first half of the past day is equal to or more than the first half of the prediction day, and a negative value if the first half of the past day is less than the first half of the prediction day. The value of the first half flag, which is a variable that takes a value, is obtained, and takes a positive value if the average of the latter half of the past day is greater than or equal to the average of the latter half of the forecast day, and a negative value if the average of the latter half of the past day is less than the latter half of the forecast day obtains the value of the late flag is a variable taking, if the product of the second half flag and the half flag is a positive value, it is determined that the large-small relationship within a predetermined period of time is not inverted, the said first half flag If the product of the second half of the flag has a negative value, it is determined that the large-small relationship is reversed within the predetermined period.

According to the present invention, it is possible to accurately and easily determine whether or not the magnitude relationship between the predicted date/time series data and the past date/time series data within a predetermined period is reversed, and time series data having similar tendencies are used for power demand prediction. It can be easily selected as time series data.

Another aspect of the present invention is the power demand prediction apparatus, wherein the similarity between the predicted date/time series data and the past date/time series data is calculated based on a difference in each corresponding time. A time-series data candidate acquisition unit that acquires the past date/time series data that exceeds a preset first threshold value as a time-series data candidate is provided, and the time-series data selection unit acquires the time-series data candidate. The time series data used for prediction of the power demand is selected from the past date and time series data acquired by the department.

According to the present invention, with respect to the time series data acquired by the determination based on the difference (distance), the time series data used for the prediction of the power demand is selected by determining the magnitude relationship within the above-described predetermined period , and thus the prediction of the power demand is performed. The accuracy can be improved.

Another aspect of the present invention is the power demand prediction apparatus, wherein the similarity between the predicted date/time series data and the past date/time series data is calculated based on a difference in each corresponding time. A time-series data candidate acquisition unit that acquires the past date/time series data that exceeds a preset first threshold value as a time-series data candidate is provided, and the time-series data selection unit acquires the time-series data candidate. Among the past date/time series data acquired by the department, those having the similarity exceeding a second threshold value (>first threshold value) are selected as the time series data used for the prediction of the power demand, and the time series data is selected. wherein said similarity of said past time-series data candidate obtaining unit has obtained about what does not exceed the second threshold value, the past time-series data Increase Decrease relationship is not inverted Te predetermined period smell, the predetermined period wherein in preference to the past time-series data Increase Decrease relationship Te smell is reversed, selects as the time series data used for the prediction of the power demand.

According to the present invention, the data whose similarity exceeds the second threshold (>the first threshold) is selected as time-series data used for prediction of power demand, and the data whose similarity does not exceed the second threshold. Selects the time series data to be used for predicting the power demand by determining the magnitude relationship within the above-described predetermined period , so that the time series data can be efficiently selected.

Another aspect of the present invention is the power demand prediction apparatus, wherein the time-series data candidate acquisition unit is a sum of square roots of a difference between the predicted date-time series data and the past date-time series data for each time. An index having a characteristic that the smaller the value is, the greater the degree of similarity is, is the similarity.

In addition, the problems disclosed by the present application and the solutions therefor will be made clear by the section of the embodiments for carrying out the invention and the drawings.

According to the present invention, it is possible to accurately predict power demand.

It is a figure which shows the main hardware with which the power demand prediction apparatus 10 is equipped. It is a figure which shows the main function with which the power demand prediction apparatus 10 is equipped, and the main information which the power demand prediction apparatus 10 memorize|stores. It is a flow chart explaining electric power demand prediction processing S300. It is an example of predicted date/time series data and past date/time series data.

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 illustrates main hardware included in the power demand prediction apparatus 10 shown as an embodiment, and FIG. 2 illustrates main functions included in the power demand prediction apparatus 10 and main information stored in the power demand prediction apparatus 10. .. The power demand prediction device 10 is configured using an information processing device (personal computer, workstation, mainframe, etc.). The power demand prediction device 10 is installed, for example, in a central power supply command center of an electric power company.

The power demand prediction device 10 predicts the power demand by performing, for example, time series analysis (Time Series Analysis), regression analysis, analysis by a neural network, or the like based on the time series data that is the actual value of the power demand. The power demand prediction device 10 selects appropriate data from the accumulated time series data as time series data used for power demand prediction, and predicts the power demand using the selected time series data.

At the time of the selection, the power demand prediction device 10 stores, for example, predicted date/time series data that is time series data in a predetermined period of the predicted date (for example, a period from the start of the predicted date (for example, midnight) to the present). , Based on the difference (difference (distance) corresponding to each time) from the past date/time series data, which is the time series data on the day before the predicted date, the degree of similarity between them (the closer the value, the larger the value) Set)), and the past date/time series data whose similarity exceeds the preset first threshold value (>0) is acquired as a candidate for the time series data used for the prediction of power demand.

Further, the power demand prediction apparatus 10 compares the tendency of the predicted date/time series data (whether it is an increasing tendency or a decreasing tendency) with the tendency of the past date/time series data for the above-described predetermined period of the predicted date, and the tendency is similar. The past date/time series data is selected as the time series data to be used for predicting the power demand, prioritizing the past date/time series data having a dissimilar tendency.

More specifically, the power demand prediction apparatus 10 compares the predicted date/time series data with the past date/time series data in the predetermined period, and determines the past date/time series data whose magnitude relationship is not reversed in the predetermined period in the predetermined period. It is selected as time-series data to be used for power demand prediction, prioritizing past date/time series data in which the magnitude relationship is reversed.

As illustrated in FIG. 1, the power demand prediction device 10 includes a processor 11, a storage device 12, an input device 13, an output device 14, and a communication device 15. The processor 11 is configured using, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). The storage device 12 is a RAM (Random Access Memory), a ROM (Read Only Memory), an NVRAM (Non Volatile RAM), a hard disk drive, an SSD (Solid State Drive), and stores programs and data.

The input device 13 is a user interface (keyboard, mouse, etc.) for inputting information. The output device 14 is a user interface that provides various kinds of information to the user, and is, for example, a liquid crystal monitor, a printing device, or the like. The communication device 15 is configured using a communication interface such as a NIC (Network Interface Card), and other devices (for example, a monitoring control system, a system control system, a distribution automation system, various database systems, a Web server on the Internet, etc.). Communicate with.

As shown in FIG. 2, the power demand prediction device 10 includes functions of a time series data storage unit 211, a time series data candidate acquisition unit 212, a time series data selection unit 213, and a power demand prediction unit 214. The power demand prediction apparatus 10 may further include a DBMS (DataBase Management System) function. These functions are implemented by the processor 11 executing a program stored in the storage device 12 or by hardware included in the power demand prediction device 10.

As shown in the figure, the power demand forecasting device 10 includes time series data 251, calendar information 271 (day of the week, holidays, seasons, etc.), weather information 272 (at least each day corresponding to the time series data 251). Of the weather information (including temperature, humidity, wind speed, solar radiation amount, etc.) and economic information 273 (economic trends for several years corresponding to the time series data 251) are stored. The power demand prediction apparatus 10 acquires the calendar information 271, the weather information 272, and the economic information 273 via, for example, the Internet, and manages these information as, for example, a database table.

Among the above functions, the time-series data storage unit 211 is a record of the power demand (power consumption amount) on each day from a certain number of years before the present to the present, which is acquired at any time by other systems operating at the central power supply command station. The time-series data 251 that is a value (for example, data acquired at a predetermined time interval such as a time unit, a minute unit, or a second unit) is stored. It should be noted that each time series data 251 is accompanied by information indicating the date and time when each was acquired.

The time-series data candidate acquisition unit 212, based on the difference (distance) at each corresponding time between the predicted date/time series data and the past date/time series data for a predetermined period of the predicted date (for example, the period until the present of the predicted date). The degree of similarity between the two is obtained, and the past date/time series data whose degree of similarity exceeds a preset first threshold value is acquired as a candidate for time series data used for prediction of power demand. As the similarity, the time-series data candidate acquisition unit 212 uses, for example, the square sum square root of the difference (distance) between the predicted date/time series data and the past date/time series data at each time. The index having the characteristic that the degree of is increased is set. The time-series data candidate acquisition unit 212 may further acquire, as candidates for the time-series data, the prediction date/time series data and the calendar information, the weather information, the economic information, or the like that are the same or similar.

As described above, the time-series data candidate acquisition unit 212 obtains the degree of similarity based on the difference (distance) between the predicted date/time series data and the past date/time series data at each corresponding time. The time-series data acquired by 212 may include past date-time series data having an increasing tendency or a decreasing tendency different from the predicted date-time series data.

The time-series data selection unit 213 compares the prediction date/time series data with the past date/time series data for a predetermined period of the prediction date (for example, the period up to the present of the prediction date), and the magnitude relationship is not reversed in the predetermined period. The date-and-time series data is selected as the time-series data to be used for predicting the power demand, prioritizing the past date-and-time series data in which the magnitude relationship is reversed in the predetermined period. As shown in the figure, the time-series data selection unit 213 includes respective functions of a period division unit 2131, an average calculation unit 2132, a flag calculation unit 2133, and a magnitude relationship determination unit 2134.

Of the above functions, the period division unit 2131 divides the predetermined period into the first half period and the second half period.

The average calculation unit 2132 is the average of the predicted date/time series data in the first half period, the predicted date/half of the predicted date/time series data in the latter half period, and the average of the past date/time series data in the first half period. An average of the first half of the day and an average of the second half of the past day, which is an average of past date/time series data in the second half period, are obtained.

The flag calculation unit 2133 sets the value of the first half flag, which is a variable that takes a positive value if the previous day first half average is greater than or equal to the predicted day first half average, and takes a negative value if the past day first half average is less than the predicted day first half average. Ask. The flag calculation unit 2133 is a variable that takes a positive value if the latter half of the past day is greater than or equal to the latter half of the prediction day, and has a negative value if the latter half of the past day is less than the latter half of the prediction day. Ask for.

The magnitude relationship determination unit 2134 determines whether the tendency of the predicted date/time series data and the tendency of the past date/time series data are similar to each other. Specifically, if the product of the first half flag and the second half flag is a positive value, the magnitude relationship determining unit 2134 does not reverse the magnitude relationship between the predicted date/time series data and the past date/time series data (the tendency is similar). If the product of the first half flag and the second half flag is a negative value, it is determined that the magnitude relationship between the predicted date/time series data and the past date/time series data is reversed (the tendency is not similar) in the predetermined period.

The power demand prediction unit 214 illustrated in FIG. 2 performs future power demand by performing, for example, time series analysis, regression analysis, analysis by a neural network, etc. based on the time series data selected by the time series data selection unit 213. Predict. In the above prediction, the power demand prediction unit 214 derives, for example, a relational expression representing a correlation between a factor affecting power demand (for example, calendar information 271, weather information 272, economic information 273) and time series data, Predict the power demand using the derived relational expression.

[Processing example]
Next, a process (hereinafter, referred to as a power demand prediction process S300) performed by the power demand prediction device 10 when predicting the power demand will be described with reference to the flowchart shown in FIG.

As shown in the figure, first, the power demand prediction device 10 acquires predicted date and time series data from the stored time series data 251 (S311).

Subsequently, the time-series data candidate acquisition unit 212 of the power demand prediction device 10 selects one or more time-series data (past date-time series data) similar to the predicted date-time series data from the stored time-series data 251. Is acquired as a candidate of time series data used for prediction of power demand (S312). In addition, the time-series data candidate acquisition unit 212 further narrows down the time-series data similar to the predicted date/time series data on the condition that the calendar information, the weather information, the economic information, etc. are the same or similar to each other. You may make it acquire as a candidate of the time series data used for prediction.

Subsequently, the time-series data selection unit 213 of the power demand prediction device 10 divides the predetermined period of the scheduled date into the first half period and the second half period (S313).

Subsequently, the time-series data selection unit 213 of the power demand prediction device 10 obtains the predicted first half average and the predicted second half average (S314).

Then, the power demand prediction apparatus 10 selects one of the past date/time series data acquired in S312 (S315).

Then, the time-series data selection unit 213 of the power demand prediction device 10 obtains the past first half average and the past day second half average for the selected past date/time series data (S316).

Subsequently, the time-series data selection unit 213 of the power demand prediction device 10 obtains the values of the first half flag and the second half flag (S317), and determines whether the product of the first half flag and the second half flag is positive or negative (S318). If the product of the first half flag and the second half flag is positive (S318: positive), the process proceeds to S319. If the product of the first half flag and the second half flag is negative (S318: negative), the process proceeds to S320.

In S319, the time series data selection unit 213 of the power demand prediction apparatus 10 selects the selected past date/time series data as time series data used for power demand prediction (for example, the time series data selection unit 213 has selected). A flag indicating the selection is set in the time series data 251).

In S320, the power demand prediction apparatus 10 determines whether or not there is past date/time series data that has not been obtained in S315 among the past date/time series data obtained in S312. If there is unacquired past date/time series data (S320: YES), the process returns to S315. When there is no unacquired past date/time series data (S320: NO), the process proceeds to S321.

In S321, the power demand prediction apparatus 10 predicts the power demand on the prediction date using the time series data selected in S319.

In the power demand prediction process S300 described above, the time series data selection unit 213 makes a determination (S315 to S320) for all the time series data acquired by the time series data candidate acquisition unit 212 in S312. , The time-series data used to predict the power demand is selected. For example, the time-series data selection unit 213 determines that the similarity is the second threshold value among the time-series data acquired by the time-series data candidate acquisition unit 212 in S312. Time-series data that exceeds (>the first threshold) is selected as time-series data used for power demand prediction without being subject to the determination (S315 to S320) by the time-series data selection unit 213, and the similarity is The processing of S315 to S320 may be performed only for the time-series data that is less than the second threshold. By doing so, the processing load of the power demand prediction device 10 can be reduced.

[Concrete example]
FIG. 4 is an example of the predicted date/time series data and the past date/time series data. In the figure, X(i) is predicted date/time series data, and Y1(i), Y2(i), Y3(i), and Y4(i) are all past date/time series data. Note that i (predetermined natural number) represents the present, i-1 to i-5 represent past time points, and i+1 to i+s (s is a predetermined natural number) represent future time points. In this example, the period from i-5 to i corresponds to the above-mentioned predetermined period.

In this example, the time-series data candidate acquisition unit 212 sets L1 to the distance between X(i) and Y1(i), L2 to the distance between X(i) and Y2(i), and L3 to X(i) and Y3( i), L4 is the distance between X(i) and Y4(i), and the distance between the predicted date/time series data and the past date/time series data (square root sum of differences (distance) at each time) is calculated from the following formula. ..

In S312 of the power demand prediction process S300 illustrated in FIG. 3, the time series data candidate acquisition unit 212 has an index in which the degree of similarity increases as the distance decreases in the distances L1 to L4 (for example, L3 is the maximum value). In this case, the index=−0.5×(L1 to L4)/L3+1) is used as the similarity, and for example, the past date/time series data in which the similarity exceeds the first threshold value among the four distances L1 to L4 is used. It is acquired as a candidate for time-series data (past date/time series data) used for power demand prediction.

Further, in S313 of FIG. 3, the period division unit 2131 of the time-series data selection unit 213 sets, for example, i-5 to i-3 in the first half period, i-2 to i in the second half period, and the predetermined period. The division is performed (S313), and the average calculation unit 2132 obtains the predicted first half average and the predicted second half average (S314), for example, as follows.

Then, the time series data selection unit 213 selects one past date/time series data acquired in S312 (S315). Then, the average calculation unit 2132 of the time-series data selection unit 213 obtains the first half of the past day and the second half of the past day as follows (S316).

Subsequently, the flag calculation unit 2133 of the time-series data selection unit 213 obtains the first half flag and the second half flag based on the following criteria (S317).
If the first half of the past day ≥ the first half of the prediction day, the first half flag = +1 (positive)
If the average of the first half of the previous day <the average of the first half of the predicted day, the first half flag = -1 (negative)
If the latter half of the past day ≥ the latter half of the prediction day, the latter half flag = +1 (positive)
If the latter half of the past day <the latter half of the forecast day average, the latter half flag = -1 (negative)

Then, when the product of the first half flag and the second half flag is a positive value, the size relationship determination unit 2134 of the time series data selection unit 213 reverses the size relationship between the predicted date and time series data and the past date and time series data in the predetermined period. If not, and if the product of the first half flag and the second half flag is a negative value, it is determined that the magnitude relationship between the predicted date/time series data and the past date/time series data is reversed in the predetermined period (S318).

The time-series data selection unit 213 selects the past date/time series data as time-series data used for prediction based on the determination result (S319). That is, the time-series data selection unit 213 selects past date-time series data whose magnitude relationship is not reversed as time-series data used for power demand prediction (S318: positive, S319), and regarding past date-time series data whose magnitude relationship is reversed. Is excluded from the time-series data used to predict the power demand (S318: negative).

Subsequently, the power demand prediction device 10 determines whether or not there is the past date/time series data that has not been obtained in S315 among the past date/time series data acquired in S312. If there is unacquired past date/time series data (S320: YES), the process returns to S315. When there is no unacquired past date/time series data (S320: NO), the process proceeds to S321.

In S321, the power demand prediction apparatus 10 predicts the power demand on the prediction date using the time series data selected in S319.

As described above, according to the power demand prediction apparatus 10 of the present embodiment, the past date/time series data whose magnitude relationship is not reversed in the predetermined period is prioritized over the past date/time series data whose magnitude relationship is reversed in the predetermined period. And select it as time-series data used for forecasting power demand. Therefore, time series data having a tendency similar to that of the prediction date can be selected as the time series data used for the prediction of the power demand. Further, it is possible to exclude time series data having different characteristics, which cannot be excluded only by the determination based on the difference (distance). Therefore, it is possible to improve the prediction accuracy of the power demand, and eventually realize economical system operation and reduce the imbalance (difference between the power consumption and the generated power).

By the way, the above description is for facilitating the understanding of the present invention and does not limit the present invention. It goes without saying that the present invention can be modified and improved without departing from the spirit thereof and that the present invention includes equivalents thereof.

For example, in the above, the case where the period dividing unit 2131 divides the predetermined period into the first half period and the second half period has been described as an example, but the predetermined period is divided into three or more and the predicted date/time series data in each period is divided. The similarity between the predicted date/time series data and the past date/time series data may be determined based on the magnitude relationship between the past date/time series data. However, since the processing load increases as the number of divisions increases, it is preferable not to increase the number of divisions more than necessary if priority is given to the speed of processing.

By setting a plurality of the predetermined time periods and comparing the predicted date/time series data and the past date/time series data for each of the plurality of predetermined time periods, the magnitude relationship described above for each of the plurality of predetermined time periods is determined, and the plurality of predetermined time periods are determined. The degree of similarity between the predicted date/time series data and the past date/time series data may be determined by comprehensively considering the determination results of the respective size relationships. However, since the processing load increases as the number of predetermined periods increases, it is preferable not to increase the predetermined period more than necessary if priority is given to the speed of processing.

The start time of the predicted day may not necessarily be midnight, and for example, the start time of the predicted day may be midnight of the previous day. Further, the end of the predicted day may not necessarily be midnight, and for example, the end of the predicted day may be midnight on the next day.

10 power demand prediction device, 211 time series data storage unit, 212 time series data candidate acquisition unit, 213 time series data selection unit, 2131 period division unit, 2132 average calculation unit, 2133 flag calculation unit, 2134 magnitude relationship determination unit, 214 Power demand forecasting unit, 251 time series data

Claims (10)

A processor and a storage device,
A time series data storage unit that stores time series data that is the actual value of the power demand,
For a predetermined period of the predicted date, the predicted date/time series data, which is the time series data in the predetermined period, is compared with the past date/time series data, which is the time series data on a day before the predicted date, and the magnitude is changed in the predetermined period the past time-series data when the relationship is not inverted, the predetermined period said in preference to the past time-series data Increase Decrease relationship is reversed Te smell, selects as the time series data used for predicting the power demand, the time series Data selection section,
A power demand forecasting unit that forecasts power demand on the forecast day based on the selected time series data,
With
Power demand forecasting device. The power demand prediction apparatus according to claim 1, wherein
The time-series data selection unit,
The predetermined period is divided into a first half period and a second half period,
Forecast date first half average, which is the average of the forecast date and time series data in the first half period,
Forecast date latter half average, which is the average of the forecast date and time series data in the latter half period,
The first half of the past day, which is the average of the past date and time series data in the first half period,
The last half of the past day, which is the average of the past date/time series data in the latter half period,
Seeking
If the previous day first half average is equal to or higher than the prediction day first half average, a positive value is taken, and if the past day first half average is less than the prediction day first half average, a negative value is obtained.
If the latter half of the past day average is greater than or equal to the latter half of the prediction day, a positive value is taken, and if the latter half of the past day is less than the latter half of the prediction day, a negative value is obtained.
If the product of the second half flag and the half flag is a positive value, it is determined that the large-small relationship within a predetermined period of time is not inverted,
If the product of the first half flag with the late flag is negative value, it is determined that the large-small relationship is reversed within the predetermined time period,
Power demand forecasting device. It is an electric power demand prediction apparatus of Claim 1 or 2, Comprising:
The similarity between the predicted date/time series data and the past date/time series data is calculated based on the difference in each corresponding time, and the past date/time series data in which the similarity exceeds a preset first threshold value is used. A time-series data candidate acquisition unit that acquires as a time-series data candidate is provided.
The time-series data selection unit selects the time-series data used for prediction of the power demand from the past date/time series data acquired by the time-series data candidate acquisition unit,
Power demand forecasting device. The power demand prediction apparatus according to claim 3,
The similarity between the predicted date/time series data and the past date/time series data is calculated based on the difference in each corresponding time, and the past date/time series data in which the similarity exceeds a preset first threshold value is used. A time-series data candidate acquisition unit that acquires as a time-series data candidate is provided.
The time-series data selection unit,
Of the past date/time series data acquired by the time series data candidate acquisition unit, data whose similarity exceeds a second threshold value (>first threshold value) is selected as the time series data used for prediction of the power demand. Then
For those the similarity of the past time-series data the time-series data candidate obtaining unit has obtained does not exceed the second threshold value, the past time-series data the predetermined period High Low relationship is not inverted Te smell , the predetermined period said in preference to the past time-series data Increase Decrease relationship is reversed Te smell, selects as the time series data used for predicting the power demand,
Power demand forecasting device. The power demand prediction apparatus according to claim 3 or 4, wherein
The time-series data candidate acquisition unit, the index having a characteristic that the degree of similarity increases as the value of the square root of the sum of squares of the differences between the predicted date-time series data and the past date-time series data decreases Degree,
Power demand forecasting device. An information processing device having a processor and a storage device,
Storing time-series data that is the actual value of the power demand,
For a given period of predicted day, the comparison between past time-series data is the time series data of the previous day than the predicted time series data and the predicted date is the time-series data in a predetermined period, Te the predetermined period smell the step of selecting the past time-series data Increase Decrease relationship is not inverted, the predetermined period said in preference to the past time-series data Increase Decrease relationship is reversed Te odor, as the time series data used for predicting the power demand ,as well as,
Predicting the power demand on the forecast day based on the selected time series data,
A power demand forecasting method for executing. The power demand forecasting method according to claim 6, wherein
The information processing device is
Dividing the predetermined period into a first half period and a second half period,
Forecast date first half average, which is the average of the forecast date and time series data in the first half period,
Forecast date latter half average, which is the average of the forecast date and time series data in the latter half period,
The first half of the past day, which is the average of the past date and time series data in the first half period,
The last half of the past day, which is the average of the past date/time series data in the latter half period,
The step of seeking
A step of obtaining a value of a first half flag which is a variable that takes a positive value if the previous day first half average is equal to or greater than the prediction day first half average, and takes a negative value if the past day first half average is less than the prediction day first half average,
A step of obtaining a value of a second half flag which is a variable that takes a positive value if the past day latter half average is equal to or higher than the prediction day latter half average, and takes a negative value if the past day latter half average is less than the prediction day latter half average,
Wherein if the product of the first half flag and the late flag is a positive value, determining that the large-small relationship within a predetermined period of time is not inverted,
The first half when the product of the flag and the late flag is negative value, determining that the large-small relationship within the predetermined time period is inverted,
A power demand forecasting method that further executes. The power demand prediction method according to claim 6 or 7, wherein
The information processing device is
The similarity between the predicted date/time series data and the past date/time series data is calculated based on the difference in each corresponding time, and the past date/time series data in which the similarity exceeds a preset first threshold value is used. Steps to acquire as candidates for time series data ,
Wherein from the last time-series data acquisition step of selecting the time series data used for predicting the power demand,
A power demand forecasting method that further executes. The power demand prediction method according to claim 8, wherein
The information processing device is
The similarity between the predicted date/time series data and the past date/time series data is calculated based on the difference in each corresponding time, and the past date/time series data in which the similarity exceeds a preset first threshold value is used. Steps to acquire as candidates for time series data,
Step about what degree of similarity among the past time-series data obtained were collected exceeds the second threshold value (> first threshold value) is selected as the time series data used for the prediction of the power demand and,
For those the similarity of the acquired past time-series data does not exceed the second threshold value, the past time-series data the large-small relationship Te predetermined period smell is not inverted, Large Small Te predetermined period smell A step of selecting as the time series data used for the prediction of the power demand, prioritizing the past date and time series data in which the relationship is reversed,
A power demand forecasting method that further executes. The power demand forecasting method according to claim 8 or 9,
The information processing device sets the similarity as an index having a characteristic that the degree of similarity increases as the value of the square root of the sum of squares of the time difference between the predicted date/time series data and the past date/time series data decreases. Step,
A power demand forecasting method that further executes.

Images