JP7544482B2 - ENERGY CONSUMPTION PREDICTION SYSTEM, ENERGY CONSUMPTION PREDICTION SUPPORT SYSTEM, ENERGY CONSUMPTION PREDICTION METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to an energy consumption prediction system, an energy consumption prediction support system, an energy consumption prediction method, and a program.

Patent Document 1 discloses a technology for making power predictions using only temperature information (maximum and minimum temperatures) and actual power consumption values over the past few weeks, in order to provide an energy consumption prediction system that can predict energy consumption using only the minimum amount of information that is easily available. Here, prediction accuracy is improved by classifying data by category. In particular, it is said that because it is possible to make category classifications according to the operating status of a store or business, such as store holidays and business days that fall on holidays, it is possible to make category classifications that are optimal for the user.

Patent No. 6134253

Conventional energy consumption prediction systems first required energy consumption information to be categorized, for example, by day of the week or business type. Conventionally, this categorization was carried out by interviewing energy consumers or by manual analysis.

However, for example, when conducting interviews, if the wrong information is given, the wrong settings will be made, and the accuracy of the information cannot be guaranteed. Also, when analyzing by hand, the results will vary depending on the analytical skill, and there is a possibility that the data will not be classified into the appropriate category.

The object of the present invention is to categorize energy consumption information more accurately in a system for predicting energy consumption than in a case where the present invention is not adopted.

To this end, the present invention provides an energy consumption prediction system including an acquisition means for acquiring a plurality of pieces of energy consumption information within a predetermined range of energy consumed by a specific consumer, an understanding means for understanding the energy consumption tendency of each of the plurality of predetermined ranges from the consumption information of the plurality of predetermined ranges, a classification means for categorizing the plurality of predetermined ranges using each of the energy consumption tendencies thus understood, and a prediction means for predicting the energy consumption of the specific consumer using the category classifications obtained by the classification means.

The acquiring means may acquire consumption information for each type of energy consumed by the specific consumer, and the predicting means may predict the specific consumer's energy consumption for each type of energy using category classification.

The prediction means may generate average consumption information by classifying and averaging a plurality of pieces of consumption information within a predetermined range using a category classification, and predict the energy consumption of the specific consumer using the average consumption information. In this case, the prediction means may predict the energy consumption of the specific consumer by correcting the average consumption information with correction data generated using the category classification.

The present invention also provides an energy consumption prediction support system including an acquisition means for acquiring multiple pieces of energy consumption information within predetermined ranges of energy consumed by a specific consumer, an understanding means for understanding the energy consumption tendency of each of the multiple predetermined ranges from the consumption information of the multiple predetermined ranges, a classification means for categorizing the multiple predetermined ranges using each of the understood energy consumption tendencies, and an output means for outputting the categorization classified by the classification means as information to be used in predicting the energy consumption of the specific consumer.

The acquisition means may acquire consumption information for each type of energy consumed by the specific consumer, and the output means may output the category classification as information to be used in predicting the specific consumer's energy consumption for each type of energy.

The acquiring means may perform pre-processing on the consumption information for the plurality of predetermined ranges to make it easier for the grasping means to grasp the energy consumption tendency for each of the plurality of predetermined ranges. In this case, the pre-processing may be a process of dividing the consumption information for each of the plurality of predetermined ranges by the maximum value of the consumption information.

The present invention also provides an energy consumption prediction method including the steps of acquiring a plurality of pieces of energy consumption information within a predetermined range of energy consumed by a specific consumer, grasping an energy consumption trend for each of the plurality of predetermined ranges from the plurality of pieces of consumption information within the predetermined ranges, categorizing the plurality of predetermined ranges using each of the grasped energy consumption trends, and predicting the energy consumption of the specific consumer using the classified categories.

Furthermore, the present invention also provides a program for enabling a computer to perform the following functions: acquire multiple pieces of energy consumption information within predetermined ranges of energy consumed by a specific consumer; grasp the energy consumption trend of each of the multiple predetermined ranges from the consumption information of the multiple predetermined ranges; categorize the multiple predetermined ranges using each of the grasped energy consumption trends; and output the classified categories as information to be used for predicting the energy consumption of the specific consumer.

According to the present invention, in a system for predicting energy consumption, energy consumption information can be categorized more accurately than in a case where the present invention is not adopted.

1 is a diagram showing an example of the overall configuration of an energy consumption prediction system according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of a hardware configuration of an energy consumption prediction server according to an embodiment of the present invention. 2 is a block diagram showing an example of a functional configuration of an energy consumption prediction server according to an embodiment of the present invention. FIG. 11 is a flowchart showing an example of an operation performed when an energy consumption prediction server generates a category classification according to an embodiment of the present invention. FIG. 5 is a diagram for explaining in detail the processing of steps 103 and 104 in the flowchart of FIG. 4. FIG. 5 is a diagram for explaining in detail the processes of steps 105 and 106 in the flowchart of FIG. 4. FIG. 5 is a diagram for explaining the details of the process of step 107 in the flowchart of FIG. 4. 11 is a flowchart showing an example of an operation performed by an energy consumption prediction server according to an embodiment of the present invention when predicting energy consumption. FIG. 9 is a diagram for explaining the details of the process of step 152 in the flowchart of FIG. 8. FIG. 9 is a diagram for explaining the details of the process of step 153 in the flowchart of FIG. 8 . FIG. 9 is a diagram for explaining the details of the process of step 154 in the flowchart of FIG. 8.

The following describes in detail an embodiment of the present invention with reference to the attached drawings.

[Overall configuration of the energy consumption prediction system]
Fig. 1 is a diagram showing an example of the overall configuration of an energy consumption prediction system 1 in this embodiment. As shown in the figure, this energy consumption prediction system 1 is configured by connecting an energy consumption prediction server 10 and consumer terminals 30a, 30b, and 30c to a communication line 80. Although the consumer terminals 30a, 30b, and 30c are shown in the figure, they may be referred to as consumer terminal 30 when no distinction is made between them. Although three consumer terminals 30 are shown, there may be four or more consumer terminals. Furthermore, the communication line 80 may be, for example, the Internet.

The energy consumption prediction server 10 is a server computer that predicts energy consumption on a future day based on past energy consumption from the consumer terminal 30 and meteorological information. The energy consumption prediction server 10 may use this prediction result to calculate an optimal operation plan and perform remote automatic control of the equipment. Although the energy consumption prediction server 10 is shown as a single computer, it may include multiple computers that share and execute multiple functions of the energy consumption prediction server 10, which will be described later. In this embodiment, the energy consumption prediction server 10 is provided as an example of an energy consumption prediction system.

The consumer terminal 30 is installed in the building of a consumer that consumes energy, such as a store or business, and manages the energy consumption by the consumer. The consumer terminal 30 also periodically transmits the energy consumption to the energy consumption prediction server 10. Furthermore, when the consumer terminal 30 receives the prediction result of the energy consumption for a certain day in the future from the energy consumption prediction server 10, it displays this. The consumer terminal 30 may be, for example, a PC (Personal Computer).

[Hardware configuration of energy consumption prediction server and terminal device]
2 is a diagram showing an example of a hardware configuration of the energy consumption prediction server 10. As shown in the figure, the energy consumption prediction server 10 includes a control unit 11 that controls the operation of the entire device, a hard disk drive 12 that stores data and the like, a communication interface 13 that realizes communication via a LAN (Local Area Network) cable or the like, an input device 14 into which information is input, and a display device 15 on which information is displayed.

The control unit 11 has a CPU 111, a ROM 112 in which basic software, BIOS, etc. are stored, and a RAM 113 used as a work area. The CPU 111 may be a multi-core. The ROM 112 may be a rewritable non-volatile semiconductor memory. The control unit 11 is what is known as a computer.

The hard disk drive 12 is a device that reads and writes data to a non-volatile storage medium that has a magnetic material applied to the surface of a disk-shaped substrate. The non-volatile storage medium may be a semiconductor memory or the like. The communication interface 13 is an interface used to connect to other devices. The input device 14 is, for example, a keyboard, a mouse, or a touch panel, and the display device 15 is, for example, a liquid crystal display or an organic EL (Electro Luminescence) display.

The control unit 11, hard disk drive 12, communication interface 13, input device 14, and display device 15 are connected via a bus 19 and signal lines not shown.

The hardware configuration example shown in FIG. 2 can also be regarded as an example of the hardware configuration of the consumer terminal 30. However, when discussing the consumer terminal 30, the control unit 11, CPU 111, ROM 112, RAM 113, hard disk drive 12, communication interface 13, input device 14, and display device 15 in FIG. 2 will be referred to as control unit 31, CPU 311, ROM 312, RAM 313, hard disk drive 32, communication interface 33, input device 34, and display device 35, respectively.

[Functional configuration of the energy consumption prediction server]
3 is a block diagram showing an example of a functional configuration of the energy consumption prediction server 10 in this embodiment. As shown in the figure, the energy consumption prediction server 10 in this embodiment includes an energy consumption pattern storage unit 21, a normalization unit 22, a clustering unit 23, a category classification generation unit 24, a category classification storage unit 25, an average consumption pattern generation unit 26, a correlation information storage unit 27, a corrected consumption pattern generation unit 28, and a corrected consumption pattern transmission unit 29.

The energy consumption pattern storage unit 21 stores the energy consumption patterns of each consumer, such as a store or business establishment, for multiple days in the past. The energy consumption patterns may be stored based on the hourly energy consumption collected from the consumer terminal 30. Here, the hourly energy consumption may be the energy consumption for any time interval, but hereinafter, it is assumed to be the energy consumption for every 30 minutes. In addition, the energy consumption pattern is assumed to be accompanied by information on the date and day of the week on which the hourly energy consumption constituting the energy consumption pattern was observed. Here, in addition to the date and day of the week information, time information may be added to analyze trends by time, but hereinafter, an example will be described in which only the date and day of the week information is added. Note that energy is not limited to these, but may be, for example, electricity, steam, cold water, hot water, etc. The energy consumption pattern storage unit 21 may store the energy consumption pattern for each of these energies.

The normalization unit 22 normalizes the energy consumption patterns stored in the energy consumption pattern storage unit 21. Specifically, first, the energy consumption patterns for the most recent few days are acquired from the energy consumption patterns stored in the energy consumption pattern storage unit 21. Next, the acquired energy consumption patterns for the few days are normalized for each day. If the energy consumption patterns stored in the energy consumption pattern storage unit 21 are directly input to the clustering unit 23, there is a risk that the energy consumption patterns may be determined to be in different categories due to the difference in the daily energy consumption even if the energy consumption patterns are similar. For example, the normalization of the daily energy consumption pattern may be performed by dividing the energy consumption at each time of the energy consumption pattern by the maximum energy consumption of that day. Alternatively, a more generalized preprocessing may be performed to make the energy consumption pattern easier to cluster, but the following description will be given assuming that the above normalization is performed. The normalization unit 22 may acquire an energy consumption pattern for each energy and perform normalization. In this embodiment, one day is used as an example of a predetermined range (time range, i.e., period), and an energy consumption pattern is used as an example of energy consumption information within a predetermined range. In addition, a normalization unit 22 is provided as an example of an acquisition means for acquiring multiple pieces of energy consumption information within a predetermined range of energy consumed by a specific consumer, and an acquisition means for performing preprocessing on the multiple pieces of consumption information within the predetermined ranges to make it easier for the acquisition means to grasp the energy consumption tendency for each of the multiple predetermined ranges.

The clustering unit 23 performs clustering of the energy consumption patterns normalized by the normalization unit 22, and assigns to each energy consumption pattern the cluster number of the cluster into which it is classified. Here, the clustering may be performed using, for example, the K-means method. If the number of clusters to be classified is not fixed, the number of clusters may be determined using an existing algorithm such as the Jain-Dubes (JD) method, and this number of clusters may be used in the K-means method, or the X-means method, which is an extension of the K-means method, may be used instead of the K-means method. In addition, since information on the day of the week is added to each energy consumption pattern, the clustering unit 23 associates the day of the week added to the energy consumption pattern with the cluster number assigned to the energy consumption pattern. The clustering unit 23 may perform clustering for each energy. In this embodiment, the energy consumption pattern for each cluster is used as an example of the energy consumption tendency for each of the multiple predetermined ranges, and the clustering unit 23 is provided as an example of a grasping means for grasping the energy consumption tendency from the consumption amount information for the multiple predetermined ranges.

The category classification generation unit 24 generates a category classification based on the clustering result by the clustering unit 23. Specifically, the number of occurrences of each cluster number is tallied for each day of the week based on the correspondence between the day of the week and the cluster number. Next, the day of the week is classified into the cluster of the cluster number with the highest number of occurrences. Next, the cluster number of the cluster into which the day of the week is classified is set as the category classification. At this time, the category classification generation unit 24 may inquire of the user whether the category classification is appropriate, and may officially adopt it as the category classification if the user responds that it is appropriate. The category classification generation unit 24 may generate a category classification for each energy. In this embodiment, the category classification generation unit 24 is provided as an example of a classification means for classifying multiple predetermined ranges into categories using energy consumption tendencies.

The category classification storage unit 25 stores the category classifications generated by the category classification generation unit 24. The category classification storage unit 25 may store category classifications for each energy.

The average consumption pattern generating unit 26 generates an average energy consumption pattern for a target day (hereinafter simply referred to as a "target day") for predicting energy consumption from the energy consumption pattern stored in the energy consumption pattern storage unit 21, using the category classification stored in the category classification storage unit 25. Specifically, first, a category to which the target day belongs is identified among a plurality of categories included in the category classification stored in the category classification storage unit 25. Next, the energy consumption for each hour is obtained from the energy consumption patterns of a plurality of days belonging to this identified category among the energy consumption patterns stored in the energy consumption pattern storage unit 21. Next, the energy consumption for each hour is averaged for the plurality of days to create an average consumption pattern corresponding to the category to which the target day belongs. The average consumption pattern generating unit 26 may generate an average consumption pattern for each energy. In this embodiment, an average consumption pattern is used as an example of average consumption information obtained by averaging consumption information within a plurality of predetermined ranges. Furthermore, among the prediction means for predicting the energy consumption of a specific consumer using category classification, an average consumption pattern generation unit 26 is provided as an example of a function for generating average consumption information by classifying and averaging consumption information within multiple predetermined ranges using category classification.

The correlation information storage unit 27 stores correlation information (e.g., a correlation diagram) showing the correlation between temperature and energy consumption for each category included in the category classification stored in the category classification storage unit 25. Here, the correlation information shows the correlation between temperature and energy consumption, but it may show the correlation between other environmental factors and energy consumption. In addition, since the correlation information is used when correcting the average consumption pattern, it may be more generalized as correction data used when correcting the average consumption pattern. Furthermore, the correlation information may be generated in advance by a user for each category included in the category classification in response to the presentation of the category classification stored in the category classification storage unit 25. Alternatively, assuming that the energy consumption pattern stored in the energy consumption pattern storage unit 21 is linked to the temperature, etc., the correlation information may be automatically generated for each category included in the category classification stored in the category classification storage unit 25 from the energy consumption pattern and the temperature, etc. Note that the correlation information storage unit 27 may store correlation information for each energy. In this embodiment, the correlation information is used as an example of correction data generated using the category classification.

The corrected consumption pattern generating unit 28 corrects the average consumption pattern generated by the average consumption pattern generating unit 26 based on the category classification stored in the category classification storage unit 25 and the correlation information stored in the correlation information storage unit 27, and generates a corrected consumption pattern. Specifically, first, the category to which the target day belongs is identified from among the multiple categories included in the category classification stored in the category classification storage unit 25. Alternatively, the category may be identified by receiving the result identified by the average consumption pattern generating unit 26. Next, correlation information corresponding to the identified category is obtained from the correlation information stored in the correlation information storage unit 27. Next, the predicted temperature of the target day is obtained, for example, from the Internet, and the predicted energy consumption of the target day is obtained based on the obtained predicted temperature and the correlation information. Next, the average consumption pattern is corrected so that the total energy consumption in the average consumption pattern and the calculated predicted energy consumption match, and a corrected consumption pattern is generated. The corrected consumption pattern generating unit 28 may generate a corrected consumption pattern for each energy. In this embodiment, among the prediction means that predicts the energy consumption of a specific consumer using category classification, a corrected consumption pattern generation unit 28 is provided as an example of a function for correcting average consumption information with correction data.

The corrected consumption pattern transmission unit 29 transmits the corrected consumption pattern generated by the corrected consumption pattern generation unit 28 to the consumer terminal 30. The corrected consumption pattern transmission unit 29 may transmit the corrected consumption pattern for each energy.

These functional units are realized by the cooperation of software and hardware resources. Specifically, the CPU 111 (see FIG. 2) loads programs that realize the normalization unit 22, clustering unit 23, category classification generation unit 24, average consumption pattern generation unit 26, corrected consumption pattern generation unit 28, and corrected consumption pattern transmission unit 29 from, for example, the hard disk drive 12 (see FIG. 2) to the RAM 113 (see FIG. 2) and executes the programs, thereby realizing these functional units. Furthermore, the energy consumption pattern storage unit 21, category classification storage unit 25, and correlation information storage unit 27 are realized, for example, by the hard disk drive 12 (see FIG. 2).

[Operation of Energy Consumption Prediction Server]
First, an operation performed by the energy consumption prediction server 10 in this embodiment when generating a category classification will be described.

Figure 4 is a flowchart showing an example of the operation of the energy consumption prediction server 10 when generating and storing a category classification.

As shown in the figure, in the energy consumption prediction server 10, the normalization unit 22 first obtains the energy consumption patterns for the most recent few days from the energy consumption patterns stored in the energy consumption pattern storage unit 21 (step 101).

Next, the normalization unit 22 normalizes the energy consumption patterns for the most recent few days acquired in step 101 (step 102). Specifically, the amount of energy consumption at each time in the energy consumption pattern for one day is divided by the maximum amount of energy consumption for that day. As a result, the amount of energy consumption at each time in the energy consumption pattern becomes a value between 0 and 1, making it easy to compare the energy consumption patterns as waveforms.

Next, the clustering unit 23 performs clustering on the energy consumption patterns normalized in step 102, and assigns a cluster number to each energy consumption pattern (step 103). Specifically, the clustering result information indicating the clustering result stores the correspondence between the multiple energy consumption patterns and the cluster numbers.

Then, the clustering unit 23 associates the day of the week added to each energy consumption pattern with the cluster number assigned to that energy consumption pattern (step 104). Specifically, in the clustering result information, the day of the week added to each energy consumption pattern is associated with that energy consumption pattern to generate a correspondence between the day of the week and the cluster number. At that time, the date added to the energy consumption pattern may also be associated with the correspondence between the day of the week and the cluster number.

Next, the category classification generation unit 24 tallies the number of occurrences of each cluster number for each day of the week based on the correspondence between the days of the week and the cluster numbers generated in step 104 (step 105).

Then, based on the number of occurrences of each cluster number for each day of the week tallied in step 105, the category classification generation unit 24 classifies the cluster number that appears most frequently for each day of the week as a category classification (step 106).

Finally, the category classification generation unit 24 stores the category classification generated in step 106 in the category classification storage unit 25 (step 107).

Here, some steps in the flowchart in FIG. 4 will be explained using concrete examples. Note that in the flowchart in FIG. 4, it is assumed that a category classification is generated to be used for predicting energy consumption, but in the following explanation, it is assumed that a category classification is generated to be used for predicting energy consumption, particularly power consumption.

Figure 5 is a diagram for explaining the processing of steps 103 and 104 in detail. Figure 5 shows an example of the clustering result information generated in steps 103 and 104.

As shown in the figure, the clustering result information associates dates, days of the week, power consumption patterns, and cluster numbers. Among these, the power consumption patterns are normalized power consumption patterns stored in the energy consumption pattern storage unit 21. The cluster numbers were assigned to the power consumption patterns in step 103. The days of the week were associated with the power consumption patterns in step 104. Note that, although the figure shows a state in which cluster numbers are assigned to power consumption patterns for the period from June 1, 2018 to June 10, 2018, in reality, cluster numbers are assigned to power consumption patterns for a longer period.

Figure 6 is a diagram for explaining the processing of steps 105 and 106 in detail. Figure 6 shows the counted results of the number of occurrences of each cluster number for each day of the week in step 105. Note that in the figure, cluster numbers 1 to 7 are shown, but this is merely an example and does not limit the number of clusters.

For example, if we look at Friday, in the period shown in Figure 5, cluster number 1 is assigned twice, as shown in rows 211 and 213. Meanwhile, in other periods, cluster number 1 is assigned 10 times, cluster number 2 is assigned once, and cluster number 4 is assigned once. Then, in Figure 6, as shown in row 221, 12 is recorded for cluster number 1, 1 for cluster number 2, and 1 for cluster number 4 for Friday.

Furthermore, if we look at Saturdays, in the period shown in FIG. 5, cluster number 2 is assigned twice, as shown in rows 212 and 214. Meanwhile, in other periods, cluster number 2 is assigned 11 times, and cluster number 3 is assigned once. Then, in FIG. 6, as shown in row 222, 13 is recorded for cluster number 2 and 1 is recorded for cluster number 3 for Saturdays.

Then, in step 106, each day of the week is classified into the cluster with the cluster number that appears most frequently. For example, Friday is classified into cluster number 1 because cluster number 1 appears most frequently at 12 times. Saturday is classified into cluster number 2 because cluster number 2 appears most frequently at 13 times. Similarly, Sunday is classified into cluster number 2, and Monday through Thursday are classified into cluster number 1. Note that in the figure, the columns for the days of the week classified into cluster number 1 and the number of occurrences of cluster number 1 for those days are hatched with diagonal lines, and the columns for the days of the week classified into cluster number 2 and the number of occurrences of cluster number 2 for those days are shaded.

Figure 7 is a diagram for explaining the process of step 107 in detail. Figure 7 shows an example of category classification stored in the category classification storage unit 25.

In FIG. 6, Monday to Friday are classified into cluster number 1, and Saturday and Sunday are classified into cluster number 2, so in FIG. 7, Monday to Friday are classified into category 1, and Saturday and Sunday are classified into category 2. In conventional methods of generating category classifications by interviewing consumers or by manual analysis, Saturday and Sunday may have been set into separate categories based on business hours, for example, but the method of this embodiment classifies categories more appropriately.

Each category is divided into day and night. The hours of day and night can be set appropriately depending on the business style of the target store, business, etc. For example, if the business hours of the target store, business, etc. are from 9:00 to 21:00, the hour before and after that can be considered day, and the other hours can be considered night. For example, day can include the hours from 12:00 p.m. to 2:00 p.m., when the temperature is usually at its highest, and night can include the hours from 3:00 a.m. to 5:00 a.m., when the temperature is usually at its lowest. This division into day and night can be set by clustering the amount of power consumption at each time, but it can also be set by interviewing consumers or by manual analysis, as in the past.

Next, we will explain the operation of the energy consumption prediction server 10 in this embodiment when predicting and transmitting energy consumption.

Figure 8 is a flowchart showing an example of the operation of the energy consumption prediction server 10 in this embodiment when predicting energy consumption. This flowchart shows an example of the operation when predicting energy consumption for a target day (e.g., tomorrow).

As shown in the figure, in the energy consumption prediction server 10, the average consumption pattern generation unit 26 first identifies the category to which the target day belongs from among the multiple categories included in the category classification stored in the category classification storage unit 25 (step 151).

Then, the average consumption pattern generating unit 26 generates an average energy consumption pattern for the target day from the energy consumption patterns for multiple days belonging to the category identified in step 151 among the energy consumption patterns stored in the energy consumption pattern storage unit 21 (step 152). Specifically, the average consumption pattern for the target day is generated by taking the average of the energy consumption amounts at each time for the multiple days in the energy consumption patterns for the multiple days belonging to the category identified in step 151.

Next, the corrected consumption pattern generation unit 28 obtains the predicted energy consumption for the target day based on the correlation information stored in the correlation information storage unit 27 that corresponds to the category identified in step 151 and the predicted temperature for the target day (step 153).

Then, the corrected consumption pattern generation unit 28 corrects the average consumption pattern generated in step 152 so that the total energy consumption in the average consumption pattern matches the predicted energy consumption obtained in step 153, thereby generating a corrected consumption pattern (step 154). Specifically, the energy consumption at each time in the average consumption pattern is multiplied by a certain constant to obtain the corrected energy consumption at each time, and the corrected consumption pattern is generated based on this.

Finally, the corrected consumption pattern transmission unit 29 transmits the corrected consumption pattern generated in step 154 to the consumer terminal 30 (step 155). As a result, the consumer terminal 30 displays the corrected consumption pattern. The consumer terminal 30 may also display the allowable energy consumption amount for the target store, business establishment, etc., and a warning when the expected energy consumption is predicted to exceed the allowable energy consumption amount.

Here, some steps in the flowchart in FIG. 8 will be explained using concrete examples. Note that while the flowchart in FIG. 8 assumes that energy consumption is predicted, the following explanation will assume that energy consumption, particularly power consumption, is predicted.

Figure 9 is a diagram for explaining the process of step 152 in detail. Figure 9 shows an example of an average consumption pattern generated in step 152.

The average consumption pattern generation unit 26 calculates, for example, the average power consumption for every 30 minutes from Monday to Friday in the past for category 1-daytime.

Here, the power consumption used to calculate the average value may be, for example, 10 to 15 days for a category with three or more days per week (in this example, five days from Monday to Friday). Also, for example, 5 to 9 days for a category with two or less days per week (in this example, two days from Saturday and Sunday). That is, the average power consumption of the same category over the past three to five weeks or so may be calculated. As a result, as shown in the figure, a graph of the average consumption pattern based on the average power consumption for each time (every 30 minutes) is created (A in the figure). Note that this is a graph for the day, but a graph for the night is also created in the same way. The average consumption pattern generation unit 26 also creates similar graphs for the day and night separately for other categories.

Figures 10(a) and (b) are diagrams for explaining the processing of step 153 in detail. Figure 10(a) shows a correlation diagram between temperature and total power consumption for each category, both during the day and at night. Figure 10(b) shows an enlarged view of the correlation diagram for category 1 - day.

Although not shown in Figure 10(a), the correlation diagram between temperature and total power consumption is created by plotting the temperature on the horizontal axis and the total power consumption during the day or night on the vertical axis, and the temperature for each day and the total power consumption during the day or night. In this case, for the day, the maximum temperature and the total power consumption during the day are plotted, and for the night, the minimum temperature and the total power consumption during the night are plotted. A correlation equation is then derived from these multiple plots, for example, by the least squares method.

When deriving the correlation equation, the degree of the correlation equation can be changed for each category. For example, categories with a large number of plots (e.g., categories with three or more days per week) can be approximated with a quadratic curve, and categories with a small number of plots (e.g., categories with two or fewer days per week) can be approximated with a straight line.

The corrected consumption pattern generation unit 28 acquires the predicted temperatures (predicted maximum temperature and predicted minimum temperature) for the day via the network. Next, the corrected consumption pattern generation unit 28 acquires the predicted total power consumption during the day and night corresponding to the obtained predicted temperatures according to the category of the target day. For example, in FIG. 10(b), the corrected consumption pattern generation unit 28 calculates the daytime total power consumption C corresponding to the maximum temperature B. In the case of night, the corrected consumption pattern generation unit 28 calculates the nighttime total power consumption corresponding to the minimum temperature.

Figure 11 is a diagram for explaining the process of step 154 in detail. Figure 11 shows the process by which the corrected consumption pattern generation unit 28 generates a corrected consumption pattern.

For example, if the predicted total daytime power consumption is 20% greater than the daytime total power consumption in the daytime average consumption pattern (A in the figure), the corrected consumption pattern generation unit 28 multiplies all power consumption for every 30 minutes by 1.2. Therefore, the hourly power consumption shifts upward overall (in the direction of arrow F in the figure). The resulting graph (D in the figure) becomes the corrected consumption pattern. That is, the area C enclosed by graph D in the daytime (E in the figure) coincides with the predicted total power consumption (C in FIG. 10(b)). The corrected consumption pattern generation unit 28 similarly generates a corrected consumption pattern for the night.

In this embodiment, the energy consumption prediction server 10 is configured to realize both the function of generating category classifications and the function of predicting energy consumption, but this is not limited to the above.

The energy consumption prediction server 10 may realize only the function of predicting energy consumption, and the function of generating category classification may be realized by another device. That is, the energy consumption pattern storage unit 21, category classification storage unit 25, average consumption pattern generation unit 26, correlation information storage unit 27, and corrected consumption pattern generation unit 28 may be realized in the energy consumption prediction server 10, and the normalization unit 22, clustering unit 23, and category classification generation unit 24 may be realized in an energy consumption prediction support system that supports the prediction of energy consumption. In this case, the category classification generation unit 24 is an example of an output means that outputs the category classification as information to be used in predicting the energy consumption of a specific consumer.

In addition, both or either of the functions of generating category classifications and predicting energy consumption in the energy consumption prediction server 10 may be realized in the consumer terminal 30.

[Effects of this embodiment]
As described above, in this embodiment, the energy consumption prediction server 10 automatically classifies past energy consumption patterns by clustering technology and uses the obtained classification results for category setting. This makes it possible to make the category classification used when classifying energy consumption patterns in the system for predicting energy consumption more appropriate than when the category classification is generated by interviewing consumers or by manual analysis.

1...Energy consumption prediction system, 10...Energy consumption prediction server, 21...Energy consumption pattern storage unit, 22...Normalization unit, 23...Clustering unit, 24...Category classification generation unit, 25...Category classification storage unit, 26...Average consumption pattern generation unit, 27...Correlation information storage unit, 28...Corrected consumption pattern generation unit, 29...Corrected consumption pattern transmission unit, 30...Consumer terminal

Claims (10)

An acquisition means for acquiring information on the amount of energy consumed by a specific consumer over a plurality of days ;
a clustering means for clustering the consumption information for the plurality of days and assigning a cluster number to each of the consumption information for the plurality of days ;
a classification means for tallying up the frequency of occurrence of a cluster number assigned to the consumption information to which each of the plurality of days of the week is added, and classifying each of the plurality of days of the week into a cluster having a cluster number that appears the most frequently, thereby classifying the plurality of days of the week into categories;
and a prediction means for predicting the energy consumption of the specific consumer on a target day using the consumption information of a day among the plurality of days that belongs to the category classification to which the target day belongs . The acquisition means acquires the consumption information for each of the energy consumed by the specific consumer,
The energy consumption prediction system according to claim 1 , characterized in that the prediction means predicts the energy consumption of the specific consumer for each of the energies on the target day . 2. The energy consumption prediction system according to claim 1, wherein the acquisition means performs a pre-processing of dividing the consumption information by a maximum value of the consumption information for each of the plurality of days . The energy consumption prediction system of claim 1, characterized in that the prediction means generates average consumption information by averaging the consumption information of days among the plurality of days that belong to the category classification to which the target day belongs, and predicts the energy consumption of the specific consumer on the target day using the average consumption information. The energy consumption prediction system of claim 4, characterized in that the prediction means predicts the energy consumption of the specific consumer on the target day by correcting the average consumption information with correction data for the category classification to which the target day belongs . An acquisition means for acquiring information on the amount of energy consumed by a specific consumer over a plurality of days ;
a clustering means for clustering the consumption information for the plurality of days and assigning a cluster number to each of the consumption information for the plurality of days ;
a classification means for tallying up the frequency of occurrence of a cluster number assigned to the consumption information to which each of the plurality of days of the week is added, and classifying each of the plurality of days of the week into a cluster having a cluster number that appears the most frequently, thereby classifying the plurality of days of the week into categories;
and an output means for outputting the category classification classified by the classification means as information to be used for predicting the energy consumption of the specific consumer on a target day . The acquisition means acquires the consumption information for each of the energy consumed by the specific consumer,
The energy consumption prediction support system according to claim 6, characterized in that the output means outputs the category classification as information used to predict the energy consumption of the specific consumer for each energy on the target day . 7. The energy consumption prediction support system according to claim 6, wherein the acquisition means performs a pre-processing of dividing the consumption information by a maximum value of the consumption information for each of the plurality of days . A step in which a computer acquires consumption information of energy consumed by a specific consumer for multiple days ;
A computer performs clustering of the consumption information of the plurality of days and assigns a cluster number to each of the consumption information of the plurality of days ;
a step of categorizing the days of the week by aggregating the frequency of occurrence of a cluster number assigned to the consumption information to which each day of the week is assigned, and classifying each of the days of the week into a cluster having a cluster number that appears the most frequently;
and a step of a computer predicting the energy consumption of the specific consumer on a target day using the consumption information for a day among the plurality of days that belongs to the category classification to which the target day belongs . On the computer,
A function for acquiring information on the amount of energy consumed by a specific consumer over multiple days ;
a function of clustering the consumption information for the plurality of days and assigning a cluster number to each of the consumption information for the plurality of days ;
a function of categorizing the days of the week by tallying up the frequency of occurrence of a cluster number assigned to the consumption information to which the day of the week is added for each of the days of the week and classifying each of the days of the week into the cluster having the cluster number that appears the most frequently;
and a program for realizing a function of outputting the classified category classification as information to be used for predicting the energy consumption of the specific consumer on a target day .