Energy consumption prediction method, energy consumption prediction system and program

Abstract

To improve the prediction accuracy of an amount of energy consumption using a method of excluding a peculiar day while suppressing a load of a user.SOLUTION: The present invention includes: acquiring the rate of change in demand by time (S102); acquiring an integrated value in which the absolute value of the difference between the rate of change and the average value is integrated on a daily basis (S104); designating the day of a maximum integrated value as an extraordinary day (a first peculiar day) (S107) when a determination value that is the maximum integrated value divided by the average of the integrated values except the maximum integrated value is a threshold value or larger, and acquiring the average value, the integrated value, the maximum integrated value, and the determination value in the case of excluding the first peculiar day from predetermined days in the past; and designating the day of a maximum integrated value as an extraordinary day (a second peculiar day) (S107) when the determination value is the threshold value or larger, and calculating a first actual value that is an actual value before excluding the first peculiar day, a second actual value after excluding the first peculiar day from the first actual value, and a third actual value after excluding the second peculiar day from the second actual value.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to an energy consumption prediction method, an energy consumption prediction system, and a program. [Background technology]

[0002] For example, Patent Document 1 discloses a configuration including a pattern creation means that distinguishes between day and night for each classified target category, and averages energy consumption information for a predetermined number of predetermined hours going back from the target energy consumption prediction date to create an average energy consumption pattern for each day and night; an energy consumption prediction means that correlates temperature information with energy consumption for each category, calculates the predicted daytime energy consumption from the predicted maximum temperature for the target prediction date, and calculates the predicted nighttime energy consumption from the predicted minimum temperature for the target prediction date; and an energy consumption correction means that compares the total energy consumption in the average energy consumption pattern with the predicted energy consumption for each day and night, and multiplies the average energy consumption pattern by a constant for each predetermined hour so that the energy consumption matches the predicted energy consumption. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-90639 Summary of the Invention [Problem to be solved by the invention]

[0004] Here, if there is variation in past actual values, this will affect the energy consumption forecast calculated based on the actual values, leading to a decrease in prediction accuracy.However, if a user identifies and excludes specific days with large variations in actual values, the burden on the user regarding energy consumption forecasts will increase.

[0005] An object of the present invention is to solve these problems, and specifically to improve the accuracy of energy consumption prediction by using a method that excludes peculiar days while reducing the burden on users. [Means for solving the problem]

[0006] With this objective in mind, the present invention provides: An energy consumption prediction method performed by a computer, comprising: A result value for energy consumption is acquired, a time change rate is calculated which is a rate of change for each predetermined hour over time for the acquired result value for each day of a predetermined number of days going back in time, an integrated value is calculated by integrating the absolute value of the difference between the time change rate and the average value of the time change rate on a daily basis, and a judgment value is calculated by dividing the maximum integrated value which is the integrated value of the largest value among the integrated values ​​by the average of the integrated values ​​other than the maximum integrated value, and whether or not the judgment value is equal to or greater than a threshold value, and if the result of the judgment is equal to or greater than the threshold value, the day with the maximum integrated value is set as a first particular day, and the above-mentioned calculation is performed from a predetermined day going back in time. The energy consumption prediction method calculates the average value, the integrated value, the maximum integrated value, and the judgment value when one specific day is excluded, performs the judgment based on the judgment value, and when the result of the judgment is equal to or greater than the threshold, designates the day with the maximum integrated value as a second specific day.The method calculates a first actual value, which is the actual value before excluding the first specific day, a second actual value obtained by excluding the first specific day from the first actual value, and a third actual value obtained by excluding the second specific day from the second actual value.The method predicts energy consumption using each of the calculated first to third actual values ​​in a predetermined manner. Here, the maximum integrated value and the judgment value are calculated when the second specific day is excluded, and if the result of the judgment based on the judgment value is equal to or greater than the threshold value, the processing for the third and subsequent specific days, which is processing for determining the day with the maximum integrated value as the third specific day, is performed until the result of the judgment becomes less than the threshold value, and if at least a third specific day exists as a result of the processing for the third and subsequent specific days, 、 The nth excluded day (n is an integer of 3 or greater) ofThe (n+1)th actual value after excluding it from the (n)th actual value is calculated in order up to the last special day, and the energy consumption prediction can be performed using the calculated (n)th actual value. In this case, the processing for the special days from the third onwards can be performed until the daily integrated value reaches three. Here, a prediction error can be calculated for each of the energy consumption predictions using the calculated actual values. In this case, a predetermined number of days going back in time for the actual values ​​and the threshold value can be set as variables within a predetermined range, and the calculation can be performed using a combination of each variable. Another aspect of the present invention that achieves the above object includes a means for acquiring a performance value for energy consumption, a means for calculating a time change rate which is a predetermined hourly rate of change over time for each of a predetermined number of days going back in time of the acquired performance value, a means for calculating an integrated value by integrating, on a daily basis, the absolute value of the difference between the time change rate and the average value of the time change rate, a maximum integrated value which is the integrated value of the largest value among the integrated values, by the average of the integrated values ​​other than the maximum integrated value, and a means for determining whether a judgment value is equal to or greater than a threshold value, and if the result of the determination is equal to or greater than the threshold value, designating the day with the maximum integrated value as a first particular day, and a means for calculating a time change rate which is a predetermined hourly rate of change over time for each of the acquired performance values ​​going back in time. the first specific day is excluded from the first actual value; the second specific day is excluded from the second actual value; and the third specific day is excluded from the second actual value. Another invention for achieving the above object includes, in an information processing device, a function for acquiring actual values ​​for energy consumption, a function for calculating a time change rate which is a rate of change for each predetermined hour over time for each day of a predetermined number of days going back in time for the acquired actual values, a function for calculating an integrated value by integrating, on a daily basis, the absolute value of the difference between the time change rate and the average value of the time change rate, a function for determining whether a judgment value which is a value obtained by dividing a maximum integrated value which is the integrated value of the largest value among the integrated values ​​by the average of integrated values ​​other than the maximum integrated value is equal to or greater than a threshold value, and if the result of the determination is equal to or greater than the threshold value, a function for determining that the day with the maximum integrated value is a first particular day, and a function for determining whether or not the day with the maximum integrated value is a first particular day. The program realizes the following functions: calculating the average value, the accumulated value, the maximum accumulated value, and the judgment value when the first specific day is excluded from a predetermined date going back, making the judgment based on the judgment value, and if the result of the judgment is equal to or greater than the threshold, designating the day with the maximum accumulated value as the second specific day; calculating a first actual value, which is the actual value before the first specific day is excluded, a second actual value, which is the first actual value after the first specific day is excluded from the first actual value, and a third actual value, which is the second actual value after the second specific day is excluded from the second actual value; and performing energy consumption prediction using a predetermined method using each of the calculated first to third actual values. [Effects of the Invention]

[0007] According to the present invention, it is possible to improve the accuracy of energy consumption prediction by excluding peculiar days while reducing the burden on the user. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a diagram illustrating a configuration example of an energy consumption prediction system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing an example of a functional configuration of an equipment control device. [Figure 3] FIG. 2 is a block diagram illustrating an example of a hardware configuration of a terminal. [Figure 4]The graphs show the relationship between cooling demand and temperature, with the vertical axis representing cooling demand (MJ) and the horizontal axis representing temperature (°C). (a) shows Case 1, and (b) shows Case 2. [Figure 5] FIG. 2 is a block diagram illustrating an example of a functional configuration of a terminal. [Figure 6] 10 is a flowchart illustrating a process of excluding a specific day from a prediction. [Figure 7] This is a graph showing energy demand over time, with the vertical axis representing energy demand (MJ) and the horizontal axis representing time. [Figure 8] 7A and 7B are diagrams for explaining steps 103 and 104 in FIG. 6, where (a) shows step 103 and (b) shows step 104. [Figure 9] FIG. 7 is a diagram for explaining step 109 in FIG. 6. [Figure 10] FIG. 7 is a diagram for explaining step 112 in FIG. 6. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 is a diagram showing an example of the configuration of an energy consumption prediction system 1 according to this embodiment. In this embodiment, in a configuration in which energy equipment 10 installed in a building such as an apartment building is controlled by an equipment control device 20, the energy consumption prediction system 1 provides energy consumption prediction information to the equipment control device 20 to ensure efficient operation of the energy equipment 10.

[0010] The energy consumption prediction system 1 includes a terminal 30 that performs energy consumption prediction. The terminal 30 acquires weather data from a terminal 40. The terminal 30 is connected to an equipment control device 20 and the terminal 40 via a network 50.

[0011] The energy facility 10 is a facility that generates electricity consumed in a building using, for example, city gas and supplies heat to the building as a result of the power generation, and achieves energy and cost savings by operating it efficiently. To this end, an equipment operation plan is transmitted from the terminal 30 to the equipment control device 20, and the equipment control device 20 controls the operation of the energy facility 10 based on the equipment operation plan.

[0012] The terminal 30 of the energy consumption prediction system 1 is realized by, for example, a computer. The terminal 30 may be configured by a single computer, or may be realized by distributed processing using multiple computers. The terminal 30 is an example of an information processing device. The prediction of energy consumption in the terminal 30 can be applied to gas consumption as well as power consumption.

[0013] Terminal 40 is a terminal of a weather data company, and is realized by, for example, a PC (Personal Computer), etc. Terminal 40 transmits weather data such as temperature to terminal 30 periodically or upon request.

[0014] The type of network 50 is not particularly limited as long as it is capable of transmitting and receiving data, and may be, for example, the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), etc. The communication line used for data communication may be wired or wireless. Furthermore, a configuration in which each device is connected via multiple networks 50 or communication lines may be used.

[0015] Next, the functional configuration of the equipment control device 20 will be described. FIG. 2 is a block diagram showing an example of the functional configuration of the equipment control device 20. The equipment control device 20 includes a transmitter / receiver 21 , a performance value measuring unit 22 , a performance value storage unit 23 , a prediction information acquisition unit 24 , and an operation control unit 25 .

[0016] The transmitting / receiving unit 21 is a network interface for connecting to the terminal 30 via the network 50 (see FIG. 1). This allows the equipment control device 20 to exchange data with the terminal 30.

[0017] The actual value measurement unit 22 measures the energy demand per unit time in the energy facility 10 as an actual value. The unit time here refers to, for example, 30 minutes, 1 hour, 4 hours, etc. The actual value is information on the amount of power consumed by a target building, store, or office for each predetermined time period for multiple days in the past, and is sometimes referred to as power consumption information.

[0018] The performance value storage unit 23 stores the performance values ​​measured by the performance value measurement unit 22, and periodically or upon request, reads the performance values ​​to the terminal 30. The read performance values ​​are transmitted to the terminal 30 by the transmission / reception unit 21.

[0019] The prediction information acquisition unit 24 acquires the energy consumption prediction information transmitted from the terminal 30 via the transmission / reception unit 21 . The operation control unit 25 controls the operation of the energy equipment 10 in accordance with the energy consumption prediction information acquired by the prediction information acquisition unit 24 .

[0020] In this embodiment, the equipment control device 20 is configured separately from the terminal 30, but the terminal 30 may be provided with the functional configuration of the above-described equipment control device 20. In other words, this is a modified example in which the terminal 30 controls the energy equipment 10.

[0021] FIG. 3 is a block diagram showing an example of the hardware configuration of the terminal 30. As shown in FIG. As shown in the figure, the terminal 30 includes a CPU (Central Processing Unit) 100a, which is a computing means, and a memory 100c, which is a main storage means. Each device also includes external devices such as a magnetic disk drive (HDD: Hard Disk Drive) 100g, a network interface 100f, a display mechanism 100d, an audio mechanism 100h, and input devices 100i such as a keyboard and a mouse.

[0022] The memory 100c and the display mechanism 100d are connected to the CPU 100a via the system controller 100b. The network interface 100f, the magnetic disk drive 100g, the audio mechanism 100h, and the input device 100i are connected to the system controller 100b via the I / O controller 100e. Each component is connected by various buses such as a system bus and an input / output bus.

[0023] The magnetic disk device 100g stores programs for implementing various functions. These programs are loaded into the memory 100c, and the CPU 100a executes processes based on these programs, thereby implementing various functions.

[0024] To achieve net-zero, which means the total of greenhouse gas or carbon dioxide emissions minus absorption and removal is zero, it is important not only to utilize renewable energy but also to thoroughly pursue energy-saving activities that reduce the amount of energy used. In recent years, there have been cases where energy management systems have been introduced that predict energy consumption such as electricity and heat, and optimally operate energy facilities based on the prediction results. In order to pursue energy conservation, it is important to make predictions with high accuracy.

[0025] Here, Figure 4 is a graph showing the relationship between cooling demand and temperature, with past performance values ​​shown as diamonds, predicted values ​​shown as squares, and actual values ​​shown as equilateral triangles. (a) shows Case 1, and (b) shows Case 2. The vertical axis of each of (a) and (b) is the cooling demand (MJ) on that day, and the horizontal axis is the maximum temperature (°C) on that day. As shown in Figure 4(a), when a scatter plot is created by plotting actual daily values, a positive correlation is observed between cooling demand and temperature. By determining this line, it is possible to derive cooling demand from predicted temperatures. In Case 1 in Figure 4(a), there is relatively little variation in the actual values, and the difference between predicted and actual values ​​is small. The performance data can be categorized by day of the week or business type.

[0026] On the other hand, in Case 2 in Figure 4(b), the actual daily values ​​contain outliers that are far removed from the majority of the other data, resulting in a larger difference between the predicted and actual values ​​than in Case 1, which does not contain outliers. In this way, even if there is a past day in the same category that has a clearly different trend from other days (a peculiar day), using that day for prediction will lead to a deterioration in accuracy.

[0027] Therefore, the energy consumption prediction system 1 according to this embodiment focuses on the presence of outliers that may cause discrepancies in demand predictions, and aims to improve prediction accuracy by excluding anomalous days within the same category from the past dates used for prediction. In other words, this embodiment makes effective use of readily available information to avoid major errors in energy consumption predictions and enables highly accurate predictions throughout the year. The specific details will be explained below.

[0028] First, the functional configuration of the terminal 30 will be described. FIG. 5 is a block diagram showing an example of the functional configuration of the terminal 30. As shown in FIG. As shown in the figure, the terminal 30 includes a transmitter / receiver 31, a weather data acquisition unit 32, a performance value acquisition unit 33, a consumption prediction advancement processing unit , an energy consumption prediction unit 35, and a display unit .

[0029] The transmitter / receiver 31 is a network interface for connecting to the equipment control device 20 and the terminal 40 via the network 50 (see FIG. 1). This allows the terminal 30 to exchange data with the equipment control device 20 and to communicate data with the terminal 40.

[0030] The weather data acquisition unit 32 acquires weather data such as temperature from the weather data company's terminal 40 and stores the acquired weather data. Temperature here includes the predicted maximum temperature and predicted minimum temperature, as well as past maximum and minimum temperatures. If the predicted maximum and minimum temperatures are changed, the latest weather data is acquired and stored.

[0031] The performance value acquisition unit 33 acquires performance values ​​stored in the performance value storage unit 23 from the equipment control device 20. The performance value acquisition unit 33 classifies the acquired performance values ​​by category and stores them.

[0032] The consumption prediction improvement processing unit 34 performs data processing to improve the accuracy of the consumption prediction by the energy consumption prediction unit 35, using the actual values ​​stored in the actual value acquisition unit 33. This data processing identifies outliers or outlier days in the actual values ​​that may lead to a decrease in prediction accuracy, and details will be described later.

[0033] The energy consumption prediction unit 35 predicts future energy consumption based on past actual values. More specifically, the energy consumption prediction unit 35 classifies the actual values ​​or power consumption information acquired and stored by the actual value acquisition unit 33 into categories, and for each category, averages the actual values ​​per unit time for each hour to create average power consumption patterns for daytime and nighttime for each day.

[0034] The energy consumption prediction unit 35 also creates a correlation between temperature and power consumption for each category, for daytime and nighttime, based on the temperature information (maximum and minimum temperatures) and actual values ​​for each day acquired and stored by the weather data acquisition unit 32. Furthermore, the energy consumption prediction unit 35 calculates the predicted power consumption for the target day from the temperature information for the target day.

[0035] The energy consumption prediction unit 35 compares the total power consumption in the above-mentioned average power consumption pattern with the above-mentioned predicted power consumption. Furthermore, the energy consumption prediction unit 35 calculates a corrected power consumption for every 30 minutes by multiplying the average power consumption pattern by a constant every 30 minutes so that the total power consumption in the average power consumption pattern matches the predicted power consumption. In addition, to calculate the corrected power consumption (predicted value), for example, if the predicted daytime power consumption is 20% greater than the total daytime power consumption based on the average daytime power consumption pattern, all power consumption amounts for every 30 minutes are multiplied by 1.2.

[0036] The display unit 36 ​​displays the obtained information, such as the date, the predicted maximum temperature, the predicted minimum temperature, the contracted power (allowable energy consumption), the power saving settings (settings for the power saving rate and the scheduled time of implementation), the display settings (settings for displaying / hiding graphs), the corrected power consumption (predicted value) and the actual value (power consumption), etc.

[0037] Next, data processing for excluding days with different operating patterns (peculiar days) from the prediction will be described. Fig. 6 is a flowchart showing the process of excluding peculiar days from predictions, and this process is performed by the actual value acquisition unit 33, the consumption prediction improvement processing unit 34, and the energy consumption prediction unit 35 of the terminal 30. Fig. 7 is a graph showing energy demand in a time series, with the vertical axis representing energy demand (MJ) and the horizontal axis representing time. Fig. 8 is a diagram explaining steps 103 and 104 of Fig. 6, where (a) shows step 103 and (b) shows step 104. Fig. 9 is a diagram explaining step 109, and Fig. 10 is a diagram explaining step 112.

[0038] [Steps 101 and 102] In the processing example shown in Fig. 6, when the actual value acquisition unit 33 acquires actual values ​​for energy consumption (step 101), the consumption prediction improvement processing unit 34 calculates the rate of change in demand by time of day (step 102). That is, in the example shown in Fig. 7, actual value 1 at midnight is 1000 MJ, and actual value 2 at 0:30 is 1500 MJ. In such a case, the rate of change in demand by time of day for the time period from midnight to 0:30 is calculated by dividing actual value 2 by actual value 1, that is, 1.5. The same calculation is performed for subsequent time periods.

[0039] [Step 103] Such a change rate of demand is a time change rate, which is a predetermined hourly rate of change over time, and is measured for each of a predetermined number of days going back in time from the acquired performance values. The predetermined number of days here can be, for example, 7 days, 10 days, 15 days, etc.

[0040] For example, in FIG. 8(a), as shown in the demand change numerical value column, change rates 1 to 3 are calculated for each of past day 1, past day 2, and past day 3. If the predetermined number of days is 7, change rates 4 to 7 are calculated in the same way. The same applies to FIG. 8(b), which will be described later. Then, the consumption prediction improvement processor 34 calculates the average value of the rate of change for each time period, as shown in the average column in FIG. 8(a) (step 103).

[0041] [Step 104] The consumption prediction improvement processor 34 then determines how much the determined rate of change deviates from the average value, and then calculates the integrated value of the deviation for each day (step 104). The deviation here refers to the absolute value of the difference between the demand change rate (time change rate) and the average change rate. For example, as shown in the deviation from average value column in FIG. 8(b), |average change rate - change rate 1| is calculated for past day 1. Furthermore, |average change rate - change rate 2| is calculated for past day 2, and |average change rate - change rate 3| is calculated for past day 3.

[0042] The integrated value of the deviation in step 104 is the integration of all time periods on a daily basis, and in FIG. 8(b), as shown in the integration value column, the deviations of each of past day 1 to past day 3 are integrated.

[0043] 8(a) and 8(b), the number of past days is three, but this is not limited to this, and there may be past days 4 and beyond in addition to past days 1 to 3. To explain further, in addition to the case where the number of past days is uniformly determined in advance, the user may be allowed to determine in advance a parameter indicating the past days.

[0044] [Step 105] The consumption prediction improvement processing unit 34 determines the day with the largest integrated value of the deviations for each day calculated in step 104 (step 105). That is, it identifies the maximum integrated value M1, which is the integrated value with the largest value of the integrated values ​​of deviations, and also identifies the day with the identified maximum integrated value.

[0045] [Step 106] Then, the consumption prediction improvement processor 34 calculates a first average value A1, which is the "average value of the change rate" when the change rate at each time on the day of the identified maximum integrated value M1 is excluded, and determines whether a determination value V1, which is the value obtained by dividing the maximum integrated value M1 by the first average value A1, is equal to or greater than a predetermined threshold value S. In other words, it determines whether M1 / A1=V1≧S is true (step 106).

[0046] [Step 107] If the above formula is true because the judgment value V1 is equal to or greater than the threshold value S (Yes in step 106), the day with the identified maximum integrated value M1, i.e., the day with the largest integrated value, is determined to be an error day, which is an example of a first peculiar day (step 107). In other words, the day with the largest integrated value is treated as the day most likely to deteriorate the prediction accuracy. In addition, when the threshold value S is set to 1 or less than 1, that is, when the threshold value S is set to 1 or more, the above formula always holds, and therefore, a missing day is always determined.

[0047] [Step 108] If this continues, even days that do not need to be excluded (days whose integrated value of deviation is not very different from other days) will be excluded from the prediction. Therefore, in this embodiment, the following process is performed. That is, it is determined whether the remaining past days, excluding the days determined to be out of range in step 107, are three days or less (step 108).

[0048] If the number of past days is more than three days (No in step 108), the process returns to step 103. That is, the process from step 103 onwards is performed, including branching processing, using the performance values ​​obtained by excluding the outlier days (an example of the first peculiar day) from the performance values ​​acquired by the performance value acquisition unit 33.

[0049] [Repeat processing] To provide additional explanation about the processing (repeated processing) from step 103 onwards that is performed the second time, if an outlier day is excluded as an example of a first anomalous day, the average value of the change rate is calculated in step 103, the accumulated value of the deviation is calculated in step 104, and furthermore, in step 105, the largest accumulated value is identified as the maximum accumulated value M2. Then, in step 106, a second average value A2, which is the "average value of the change rates" when the change rates at each time on the day of the maximum integrated value M2 are excluded, is calculated, and it is determined whether the judgment value V2 (= M2 / A2) is equal to or greater than the threshold value S (V2≧S). In step 107, the day of the maximum integrated value M2 is determined to be an outlier day, which is an example of a second peculiar day.

[0050] Further, to provide additional explanation regarding the case where the processing (repeated processing) from step 103 onwards is performed for the third time, if an outlier day as an example of a second special day is excluded, the average value of the change rate is calculated in step 103, and the accumulated value of the deviation is calculated in step 104, and the largest accumulated value is identified as the maximum accumulated value M3 in step 105. Then, in step 106, a third average value A3, which is the "average value of the change rates" when the change rates at each time on the day of the maximum integrated value M3 are excluded, is calculated, and it is determined whether the judgment value V3 (= M3 / A3) is equal to or greater than the threshold value S (V3≧S). In step 107, the day of the maximum integrated value M3 is determined to be an outlier day, which is an example of a third peculiar day.

[0051] The process from step 103 onwards is repeated until the determined judgment value becomes less than the threshold value S. The process from step 103 onwards is repeated until the daily integrated value reaches three.

[0052] The process of determining the outlier day, which is an example of such a third special day, is an example of "processing related to the third and subsequent special days." If there is at least a third special day as a result of the "processing related to the third and subsequent special days," the nth (n is an integer greater than or equal to 3) excluded day is excluded from the nth actual value, and the (n+1)th actual value is calculated in order up to the last special day.

[0053] Note that days defined as irregular days, which are examples of the first, second, and third peculiar days, are days on which the rate of change by time of day differs, regardless of the magnitude of the load.

[0054] [Step 109] The process of returning to step 103 is repeated until the judgment value falls below threshold value S or the number of remaining past days is three. That is, in step 106 from the first time onwards, if the judgment value is less than threshold value S (No in step 106), or if the number of remaining past days is three or less after excluding the day determined to be an incorrect day (Yes in step 108), the consumption prediction improvement processing unit 34 ranks the excluded past days in the order in which they were excluded (step 109).

[0055] More specifically, a bad day, which is an example of a first special day, is ranked 1, a bad day, which is an example of a second special day, is ranked 2, and a bad day, which is an example of a third special day, is ranked 3. For example, in the example shown in Figure 9, of past days 1 to 15, past day 2 is ranked 4, past day 3 is ranked 1, and past day 15 is ranked 8. Also, in the example shown in Figure 9, past days 1 and 4 are not considered to be bad days. Note that past day 3 in this case is an example of a first peculiar day.

[0056] The higher the ranking, the more different the operation pattern is from other days. Since demand forecasting requires that at least three days of past data be kept, the three days with the most similar operation patterns may not be ranked.

[0057] [Steps 110 and 111] After step 109, the energy consumption prediction unit 35 performs an energy consumption prediction (demand prediction) without any outlier days (step 110), then removes outlier days one by one, starting with the highest ranked, and performs an energy consumption prediction each time (step 111). That is, first, an energy consumption prediction is performed using actual values ​​without any outlier days, then an energy consumption prediction is performed using actual values ​​excluding the outlier day ranked 1 (first particular day), and then an energy consumption prediction is performed excluding the outlier day ranked 1 (first particular day) and the outlier day ranked 2 (second particular day). This type of energy consumption prediction is performed each time until the final particular day.

[0058] The actual value without any outlier days is an example of a first actual value, the actual value obtained by excluding the first special day from the actual value without any outlier days is an example of a second actual value, and the actual value obtained by excluding the first special day and the second special day from the actual value without any outlier days (the actual value obtained by excluding the second special day from the second actual value) is an example of a third actual value.

[0059] More specifically, the energy consumption prediction is performed until there are three past days that have not been excluded. In the example of past days 1 to 15 shown in Fig. 9, energy consumption prediction is performed without any excluded days, and energy consumption prediction is performed when each of the rankings 1 to 12 is excluded in turn, for a total of 13 energy consumption predictions.

[0060] [Step 112] After step 111, each time the energy consumption prediction unit 35 predicts energy consumption, it calculates a prediction error and compares it with each result. The prediction result with the highest prediction accuracy (average value of the absolute error from the actual performance) is adopted. That is, the prediction accuracy with the highest prediction accuracy is presented to the display unit 36 (see FIG. 5) (step 112).

[0061] For example, in the example shown in FIG. 10, prediction 1 and prediction 2 are shown by broken lines. Also, the actual values from 0:00 to 04:00 on the same day at the time of 04:00 are shown by a solid line. The dashed-dotted line is the actual value after 04:00 on the same day. When comparing the prediction error G1 of prediction 1 and the prediction error G2 of prediction 2, the prediction error G2 of prediction 2 is smaller than the prediction error G1 of prediction 1 (G2 < G1). Therefore, assuming that prediction 2 has a higher prediction accuracy than prediction 1, prediction 2 is adopted as the prediction result.

[0062] In the example shown in FIG. 10, it was explained that when the energy consumption prediction unit 35 takes prediction 2 as the prediction result, but it is not limited to this. The image of FIG. 10 (excluding the dashed-dotted line) may be displayed on the display unit 36 (see FIG. 5) so that the user can select prediction 1 or prediction 2.

[0063] Note that even when making a prediction for the same day, there may be cases where there is no actual performance for all time intervals (frames). In such cases, the prediction error is calculated using only the frames for which actual performance already exists.

[0064] Here, in the energy consumption prediction unit 35, the number of past days of the actual value and the threshold value S may be set as variables within a previously specified range, and implemented with each combination of the variables. In such cases, the user selects one of the combinations of these variables. The previously specified range for the number of past days is, for example, 7 to 30 days, and it is conceivable to use a one-day increment. In the example shown in FIG. 9, the number of past days is 15 days. Also, the previously specified range for the threshold value S is, for example, 1 to 1.2, and it is conceivable to use a 0.1 increment. The number of days in the past is an example of a predetermined number of days going back in time for the performance value.

[0065] Here, the performance value acquiring unit 33 of the terminal 30 is an example of a means (function) for acquiring a performance value. The consumption prediction improvement processing unit 34 is an example of a means (function) for calculating the rate of change over time, an example of a means (function) for calculating the rate of change over time, an example of a means (function) for determining the first special day, an example of a means (function) for determining the second special day, and an example of a means (function) for calculating. The energy consumption prediction unit 35 is an example of a means (function) for predicting energy consumption.

[0066] A program for realizing an embodiment of the present invention may be provided in a state stored on a computer-readable recording medium such as a magnetic recording medium (such as a magnetic tape or a magnetic disk), an optical recording medium (such as an optical disk), a magneto-optical recording medium, a semiconductor memory, etc. It may also be provided via a communication means such as the Internet.

[0067] Although the embodiments of the present invention have been described above, the technical scope of the present invention is not limited to the scope described in the above embodiments. It is clear from the claims that various modifications and improvements to the above embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0068] 1... energy consumption prediction system, 30... terminal, 33... actual value acquisition unit, 34... consumption prediction advancement processing unit, 35... energy consumption prediction unit

Claims

1. A computer-implemented energy consumption prediction method, comprising: Obtain actual values ​​for energy consumption, calculating a time change rate, which is a rate of change over time for each predetermined hour for the acquired actual value for each day of a predetermined number of days going back in time; calculating an integrated value by integrating the absolute value of the difference between the time change rate and the average value of the time change rate on a daily basis; a judgment value obtained by dividing the largest integrated value among the integrated values ​​by the average of the integrated values ​​other than the maximum integrated value is determined to be equal to or greater than a threshold value; and if the result of the judgment is equal to or greater than the threshold value, the day with the largest integrated value is designated as a first peculiar day; The average value, the integrated value, the maximum integrated value, and the judgment value are calculated from a predetermined date going back in time, excluding the first specific day, and the judgment is performed based on the judgment values. If the result of the judgment is equal to or greater than the threshold value, the day with the maximum integrated value is determined as the second specific day. calculating a first actual result value that is the actual result value before excluding the first particular day, a second actual result value that is the first actual result value after excluding the first particular day, and a third actual result value that is the second actual result value after excluding the second particular day; An energy consumption prediction method for predicting energy consumption using a predetermined method by using each of the calculated first to third actual value.

2. The maximum integrated value and the judgment value are calculated when the second specific day is excluded, and if the result of the judgment based on the judgment value is equal to or greater than the threshold value, the day with the maximum integrated value is designated as the third specific day. This processing for the third and subsequent specific days is performed until the result of the judgment becomes less than the threshold value. If there is at least a third special day as a result of the processing for the third and subsequent special days, calculate the (n+1)th actual value after excluding the nth (n is an integer of 3 or more) excluded day from the nth actual value, in order up to the last special day; The energy consumption prediction method according to claim 1 , wherein the energy consumption prediction is performed using the calculated nth actual value.

3. The energy consumption prediction method according to claim 2 , wherein the process for the third and subsequent peculiar days is performed until the daily integrated value reaches three.

4. The energy consumption prediction method according to claim 1 , further comprising: calculating a prediction error for each of the energy consumption predictions using the calculated actual value.

5. The energy consumption prediction method according to claim 4 , wherein a predetermined number of days going back in time for the actual value and the threshold value are variables within a predetermined range, and the method is carried out using a combination of the respective variables.

6. a means for obtaining actual values ​​of energy consumption; a means for calculating a time change rate, which is a rate of change over time for each predetermined hour of the acquired actual value for each day going back a predetermined number of days; means for calculating an integrated value by integrating, on a daily basis, the absolute value of the difference between the time change rate and the average value of the time change rate; a means for determining whether a judgment value, which is a value obtained by dividing the largest accumulated value among the accumulated values ​​by the average of the accumulated values ​​other than the maximum accumulated value, is equal to or greater than a threshold value, and if the result of the judgment is equal to or greater than the threshold value, designating the day with the maximum accumulated value as a first peculiar day; a means for calculating the average value, the integrated value, the maximum integrated value, and the judgment value when the first specific day is excluded from a predetermined date going back in time, making the judgment based on the judgment values, and when the result of the judgment is equal to or greater than the threshold value, determining the day with the maximum integrated value as the second specific day; means for calculating a first actual result value which is the actual result value before excluding the first particular day, a second actual result value which is the first actual result value after excluding the first particular day, and a third actual result value which is the second actual result value after excluding the second particular day; a means for predicting energy consumption by a predetermined method using each of the calculated first to third actual values; An energy consumption prediction system comprising:

7. In the information processing device, A function to obtain actual values ​​for energy consumption; a function of calculating a time change rate, which is a rate of change over time for each predetermined hour of the acquired actual value for each day of a predetermined number of days going back in time; a function of calculating an integrated value obtained by integrating, on a daily basis, the absolute value of the difference between the time change rate and the average value of the time change rate; a function of determining whether a judgment value, which is the maximum accumulated value, which is the accumulated value of the largest value among the accumulated values, divided by the average of the accumulated values ​​other than the maximum accumulated value, is equal to or greater than a threshold value, and if the result of the judgment is equal to or greater than the threshold value, determining the day with the maximum accumulated value as the first peculiar day; a function of calculating the average value, the integrated value, the maximum integrated value, and the judgment value when the first specific day is excluded from a predetermined date going back in time, making the judgment based on the judgment value, and determining the day with the maximum integrated value as the second specific day when the result of the judgment is equal to or greater than the threshold value; a function of calculating a first actual result value, which is the actual result value before excluding the first specific day, a second actual result value, which is the first actual result value after excluding the first specific day, and a third actual result value, which is the second actual result value after excluding the second specific day; a function of performing energy consumption prediction using a predetermined method by using each of the calculated first to third actual value; A program to make this happen.

Images