JP6625839B2 - Load actual data determination device, load prediction device, actual load data determination method, and load prediction method - Google Patents

Description

  The embodiments of the present invention relate to a load actual data determination device, a load prediction device, a load actual data determination method, and a load prediction method.

  2. Description of the Related Art Conventionally, in an operation system for optimally operating energy devices, in order to optimize the operation, it is necessary to increase the prediction accuracy of a load prediction device that predicts the load of the energy devices. The load prediction in the load prediction device is performed based on load result data obtained by periodically collecting load result data of energy devices to be managed. This actual load data has a day pattern that shows a similar change for each day of the day between different days, but this day pattern fluctuates due to various factors. Specifically, the daily pattern changes due to inter-annual fluctuations and seasonal fluctuations due to climate throughout the year. In addition, in order to accurately predict the fluctuation of the actual load data, it is necessary to reflect the latest actual load data as learning data or data to be subjected to statistical processing in the prediction model.

  As described above, the fluctuation of the day pattern is caused by the operation status of the equipment as a load (the activity state of the person using the equipment), the air conditioner operation status by the weather condition, and the like. Among them, differences in day types such as holidays (closed days, etc.) and weekdays (business days, working days, etc.) depending on the activities of the personnel are a major factor of fluctuation. For this reason, when performing load prediction and the like, different prediction models are often used for holidays and weekdays. In some cases, the day pattern may be further divided into day types (weekdays after holidays, Saturdays, etc.). In addition, a special (special) day is set as a special day type, and may be excluded from prediction targets. Which model is used for prediction is based on a day type such as a holiday, a weekday, or a unique day which is set in advance in calendar information for each day of the week. In addition, the setting of holidays and the setting of special days have been manually input by humans.

  In the conventional load prediction device, if the information of the day type (calendar information) of the prediction target date is incorrect, the model used is different, and the prediction result may not match the actual load pattern. In addition, since the day type of the day pattern of the past load actual data uses the calendar information at the time of the prediction as it is, data that does not match the day type of the actual day pattern may be included. By performing learning and statistical processing using such actual load data, there is a possibility that the accuracy of the prediction result is significantly reduced.

Japanese Patent No. 2704026 JP 2013-5465 A Japanese Patent No. 3160693

  The problem to be solved by the present invention is to provide a load actual data discriminating device, a load predicting device, a load actual data discriminating method, and a load predicting method capable of correcting or excluding disturbance data from load actual data. is there.

  The load result data determination device of the embodiment has a similarity acquisition unit and a type determination unit. The similarity acquisition unit acquires a similarity between a plurality of load patterns based on a load pattern indicating a transition of a load for a predetermined period obtained by measuring a load to be managed at predetermined time intervals. The type determining unit determines the type of the predetermined period in which the load pattern was obtained based on the similarity obtained by the similarity obtaining unit.

FIG. 1 is a diagram illustrating a configuration example of an operation system including a load prediction device according to a first embodiment. The figure which shows the correspondence of the data accumulated in the load prediction apparatus 10. The figure which shows the probability distribution of similarity SI, and the normal distribution of the logarithm of similarity SI. FIG. 9 is a diagram showing a specific example 1 of a load pattern for explaining a correction process of a day type in a day type determination unit 104. FIG. 9 is a diagram showing a specific example 2 of a load pattern for explaining a correction process of a day type in the day type determination unit 104. The flowchart which shows operation | movement regarding the discrimination of the day type in the load prediction apparatus 10. The figure which shows the example 3 of a load pattern for demonstrating the process which the abnormality detection part 108 detects an abnormal value. The figure which shows the specific example 4 of the load pattern for demonstrating the process which the abnormality detection part detects an abnormal value. The figure which shows the example of correction which the abnormality detection part 108 of the load prediction apparatus 10 corrected the abnormal value of the load pattern. FIG. 6 is a diagram illustrating a configuration example of a load prediction device according to a second embodiment. The figure which shows the example of the deviation in a load pattern.

  Hereinafter, a load actual data determination device, a load prediction device, a load actual data determination method, and a load prediction method according to an embodiment will be described with reference to the drawings.

(1st Embodiment)
First, an outline of an operation system including the load prediction device according to the first embodiment and optimally operating energy devices to be managed will be described.
FIG. 1 is a diagram illustrating a configuration example of an operation system including the load prediction device according to the first embodiment. As shown in FIG. 1, the operation system 1 includes a load prediction device 10, a calendar information recording unit 11, an input / output unit 12, a load performance collection unit 13, a weather information collection unit 14, a device planning unit 15, , An operation pattern recording unit 16 and a device control unit 17. The operation system 1 is connected to a load facility 2 that measures the load of the energy equipment to be managed, a weather information providing facility 3 that provides weather information, and an energy facility 4 that includes the energy equipment to be managed. I have.

  The operation system 1 determines the operation status of the energy devices to be managed based on the activity status of the staff, such as holidays (holidays, etc.), weekdays (business days, working days, etc.), and unusual days (business hours operating at different times from weekdays). , Etc.), the load is predicted according to the characteristics of each day type, and the load is managed.

  The load prediction device 10 has a function of receiving actual load data from the actual load collection unit 13 and accumulating the actual load data for a predetermined period (for example, two to three weeks). Further, the load prediction device 10 determines the day type of the load pattern based on the accumulated load result data, and determines whether the determined day type matches the day type set in the calendar information. With function. Further, the load prediction device 10 has a function of outputting a load prediction pattern in which a load pattern for each day is predicted based on the determined past load pattern for each day type. The configuration of the load prediction device 10 will be described later in detail. Note that the load pattern is a day pattern indicating a change in the load result data for each time period of one day, and the load result data itself in units of one day may be used as the load pattern.

  The calendar information recording unit 11 stores, for each date, information on the day of the week and information indicating a day type of a holiday, a weekday, or a specific day set according to the operation status of the energy appliance to be managed. Is recorded. The information on the day type is set, for example, by the administrator of the operation system 1 that operates the input / output unit 12. Specifically, for example, a setting is made such that Saturdays, Sundays, and holidays are holiday day types, and Monday to Friday excluding holidays are weekday day types. Further, a setting different from the above setting can be made according to the operation status of the energy device to be managed. For example, when the energy device is operated only on a specific day for half a day, that day is set as the day type of the specific day. The input / output unit 12 includes an input unit including a keyboard and a mouse for a user of the operation system 1 such as an administrator to perform an input operation, and an output unit including a display and a speaker for transmitting information to the user. And

  The load result collection unit 13 collects load result data in each time zone from the load equipment 2 and outputs the data to the load prediction device 10 as load result data. The weather information collection unit 14 collects weather information including temperature, humidity, rainfall, and the like from the weather information providing facility 3. The device planning unit 15 generates an operation pattern including information on an operation plan of the energy device based on the load prediction pattern recorded in the load prediction pattern recording unit 113 in the load prediction device 10. The operation pattern recording unit 16 records the operation pattern generated by the device planning unit 15. The device control unit 17 refers to the operation pattern recorded in the operation pattern recording unit 16 and controls the energy devices included in the energy equipment 4 according to the operation pattern.

  With the above configuration, the operation system 1 can appropriately manage the load of the energy device to be managed. Although the power load is managed as the load in the present embodiment, the heat load may be managed, or both the power load and the heat load may be managed.

  Next, details of the load prediction device 10 according to the first embodiment will be described. As shown in FIG. 1, the load prediction device 10 includes a load actual data recording unit 101, a similarity calculation unit 102, a daily similarity table 103, a day type determination unit 104, a determination result storage unit 105, a day type Information 106, a prediction load pattern extraction unit 107, an abnormality detection unit 108, a change notification unit 109, a prediction load pattern recording unit 110, a weather information recording unit 111, a load pattern prediction unit 112, a load prediction And a pattern recording unit 113. Although the actual load data discriminating apparatus according to the first embodiment is not shown, for example, the actual load data recording unit 101, the similarity calculating unit 102, the daily similarity table 103, and the day type determining unit 104 It is a configuration provided with:

  The load result data recording unit 101 receives and records the load result data of each time zone from the load result collection unit 13. The similarity calculating unit 102 refers to the actual load data from the actual load data recording unit 101, calculates the similarity of the load pattern between the target date and each day of the target period based on the actual load data, and calculates the daily similarity. The table 103 is created. The daily similarity table 103 is a table that stores information regarding the similarity of the load pattern between the target date and each day of the target period, generated by the similarity calculation unit 102.

  The day type determining unit 104 has a function of referring to the actual load data of the target day from the actual load data recording unit 101 to determine the day type (holiday, weekday, special day) of the load pattern of the target day. In addition, the day type determination unit 104 refers to the day type of the target day from the calendar information recording unit 11 and outputs the date type as it is to the determination result recording unit 105 if the same as the day type of the determination result. In addition, the day type determination unit 104 refers to the day type of the target date from the calendar information recording unit 11 and corrects the referenced day type to the day type of the determination result if different from the day type of the determination result, The result is output to the determination result recording unit 105. The determination result recording unit 105 records the day type of the load pattern of the target day output by the day type determination unit 104 in association with the date of the target day.

  The day type information 106 stores a statistical value of a load pattern for each day type. The statistical value of the load pattern for each day includes, for example, an average load pattern that is the average value of the load patterns for each day, and the maximum value and the minimum value of the load pattern for each day. The day type determination unit 104 calculates the statistical value of the load pattern for each day type and outputs the calculated value to the day type information 106.

  The prediction load pattern extraction unit 107 refers to the information of the day type from the determination result recording unit 105 and refers to the actual load data from the actual load data recording unit 101 to predict the load pattern from which the day type of the target period is extracted. Load pattern.

  The abnormality detecting unit 108 refers to the statistical value of the day type from the day type information 106, refers to the information of the day type of the target day from the determination result recording unit 105, and loads the load actual data of the target day from the load actual data recording unit 101. , The presence or absence of an abnormal value for each time zone in the load pattern of the target day is detected based on the statistical value corresponding to the day type of the target day. When detecting the abnormal value, the abnormality detecting unit 108 corrects the abnormal value of the load pattern, and outputs the corrected load pattern to the prediction load pattern recording unit 110.

  The change notifying unit 109 compares the day type recorded in the discrimination result recording unit 105 with the day type set in the calendar information recording unit 11 and determines whether the two date types are different. Notifies the input / output unit 12 that the day type has been changed. This allows the user of the operation system 1 to know that there is an error in the setting of the date type of the calendar information.

  The prediction load pattern recording unit 110 stores the prediction load pattern extracted by the prediction load pattern extraction unit 107. Further, when receiving the corrected load pattern from the abnormality detecting unit 108, the prediction load pattern recording unit 110 updates the load pattern at that date and time with the corrected load pattern.

  The weather information recording unit 111 records the weather information collected by the weather information collection unit 14 in association with a date. The load pattern prediction unit 112 refers to the prediction load pattern from the prediction load pattern recording unit 110, refers to the weather information from the weather information recording unit 111, and calculates the load pattern using a neural network method, a regression model, or the like. Predict and output the load prediction pattern. The load prediction pattern recording unit 113 records the load prediction pattern output by the load pattern prediction unit 112.

  Next, a specific example of the accumulation process of the daily similarity table 103 in the load prediction device 10 will be described. Further, the contents of data respectively recorded in the calendar information recording unit 11, the actual load data recording unit 101, the daily similarity table 103, and the determination result recording unit 105 included in the load prediction device 10, and the content of each data The correspondence will also be described.

FIG. 2 is a diagram illustrating a correspondence relationship of data stored in the load prediction device 10.
As shown in FIG. 2, the calendar information recording unit 11, the day type of information in each day of D ₁ day ... D S _{+ 2} days... Are set. The load prediction unit 10 has started the accumulation of load actual data _a day D. As shown on line of D ₁ day, the load actual data recording unit 101 are recorded load pattern (day pattern) corresponding to the _day D, the discrimination result in the recording unit 105 of the determination result corresponding to ₁ day D The day type is recorded. The day after D ₁ day, as shown in row ₂ days D, the load actual data recording unit 101 are recorded load pattern corresponding to ₂ days D, the discrimination result recording unit 105 corresponding to ₂ days D the determination result of the day type is recorded, day information of the similarity of the load pattern of D ₁ day and D ₂ days to similarity table 103 is recorded.

The day after D ₂ days, as shown in the row of the D ₃ days, the load actual data recording unit 101 are recorded load pattern corresponding to ₃ days D, the discrimination result recording unit 105 corresponds to ₃ days D The day type of the determination result is recorded, and the daily similarity table 103 records information on the similarity between the load patterns on the _first and _third days and information on the similarity between the load patterns on the _second and _third days. You.

  As for the day type of the determination result recorded in the determination result recording unit 105, the day type of the calendar information is temporarily set and recorded for the first few days. Since the determination of the day type by the day type determination unit 104 starts when the number of days has reached a predetermined number of days or more, the day type of the calendar information is provisionally set until then. Here, the predetermined number of days is, for example, a number of days equal to or more than the target date + 3 days, assuming that a day requiring correction is included.

The following Similarly D ₄ days, are treated with D ₅ days.., Until S days of interest period, information of the similarity to the discrimination result recording unit 105 is added one by one day. The S after the date, for example, in the line of the D _{S + 1} day, starting from the degree of similarity between _two days D, after the table ending with similarity between D _S date, a row of for example D _{S + 2} days, similar to the ₃ days D The table starts from the degree and ends with the similarity to the DS _{+ 1} day. That is, the day similarity table 103 shifts the target days included in the target period within the range of S days of the target period.

ただし、（式１）において、ｘｄ（ｔ）は、ｄ日のｔ時間帯の契約電力量Ｗｃ［ｋＷｈ］で正規化した負荷であり、Ｗｄ（ｔ）は、ｄ日のｔ時間帯の実績電力量［ｋＷｈ］である。また、（式２）において、ＳＩ_ｉｊは、ｉ日とｊ日の負荷パターンの類似度であり、ｉ日のｔ時間帯とｊ日のｔ時間帯の負荷の差の二乗の積算値により求まる値である。また、時間帯ｔ＝１，…，ｍであり、管理する時間帯を３０分間とすると、１日は２４時間なので、ｍ＝４８となる。 Next, a specific example of calculation of the similarity of the load pattern by the similarity calculator 102 will be described.
The similarity calculation unit 102 calculates the similarity SI of the load pattern for each management time period (30 minutes or 1 hour, etc.) between the target days using the following equations (Equation 1) and (Equation 2). I do.

Here, in (Equation 1), xd (t) is the load normalized by the contracted electric energy Wc [kWh] in the t time period on the d day, and Wd (t) is the actual result in the t time period on the d day. The amount of power [kWh]. In (Equation 2), SI _ij is the similarity between the load patterns on the i-th day and the j-th day, and is determined by the integrated value of the square of the difference between the loads on the t-hours on the i-th day and the t-hours on the j-th day. Value. If the time zone is t = 1,..., M, and the time zone to be managed is 30 minutes, m is 48 because one day is 24 hours.

Next, a first determination method in the day type determination using the similarity SI in the day type determination unit 104 will be described.
The similarity SI obtained by the above (Equation 2) has a positive probability distribution biased toward zero. Therefore, as shown in FIG. 3, it is assumed that the probability distribution of the similarity SI is a lognormal distribution f (x), and the normal distribution f (log (x)) of the logarithm of the similarity SI can be used. FIG. 3 is a diagram showing a probability distribution of the similarity SI and a normal distribution of the logarithm of the similarity SI.

Therefore, the day type determination unit 104 obtains the average value SIg _log対 of the logarithm of the degree of similarity between the load patterns of the same day type using the following expression (Expression 3) (the actual ￣ is (Expression 3) Above "SIg _log "). Further, the day type determination unit 104 obtains the sample standard deviation SSDg _log of the load pattern in the same day type using the following equation (Equation 4).

Further, the day type determination unit 104 uses the following (Equation 5) to calculate the average of the degree of similarity from the load pattern i to the same day type as the day type within 3σ (SSDg _log ) from the average in the day type group. If so, it is determined that they belong to the same day type. The value of σ is not limited to 3.0 shown in (Equation 5), and for example, a value between 2.5 and 3.0 may be used.

Further, a second determination technique in the day type determination using the similarity SI in the day type determination unit 104 will be described.
As shown in FIG. 2, the day similarity is stored in the day similarity table 103 in the range of the target period S days. Then, the similarity SI for each day type that is not a unique day tends to be biased toward the similarity distribution for the same day type as the target day and the similarity distribution for the other day types. Therefore, the day type determination unit 104 determines the day type using the intermediate value between the minimum value and the maximum value of the similarity SI.

First, the day type determination unit 104 sorts the similarities in the order of magnitude and sets them as [similarity 1, similarity 2,..., Similarity S]. Next, the day type discriminating unit 104 compares the sorted similarity with the intermediate value between the maximum similarity S and the minimum similarity 1 in ascending order of similarity. To find the similarity I closest to the intermediate value. That is, the similarity I is a value that satisfies the following equation, and the right side of the following equation indicates an intermediate value.
Similarity I <(Similarity S−Similarity 1) / 2

Next, the day type determination unit 104 determines, for [similarity 1, similarity 2,..., Similarity I], the same day type is equal to or greater than the set ratio and the similarity having the next magnitude to the similarity I When J satisfies the condition shown in the following (Equation 6), the day type to be determined is determined to be the same day type as the day type of the similarity 1, similarity 2,... .

  In the expression for calculating the similarity I, the denominator for obtaining the intermediate value is 2, but a value between 1.5 and 2 may be used according to the season or the like. Further, the set ratio may be changed according to the season or the like. Further, the day type determining unit 104 refers to the day type of the target day from the calendar information recording unit 11 and, if the referenced day type is different from the day type of the determination result, determines the referenced day type as the day type of the determination result. Is performed.

Next, the correction process of the day type (correction process from weekday to holiday) in the day type determination unit 104 will be described using a specific example 1 of the load pattern.
FIG. 4 is a diagram illustrating a specific example 1 of the load pattern for explaining the correction processing of the day type in the day type determination unit 104. In FIG. 4, the vertical axis is the load amount, and the horizontal axis is the time zone.

  As shown in FIG. 4, the load patterns on weekdays 1 to 5 are similar among the load patterns. The load patterns of holidays 1 to 3 are also similar among the load patterns. However, the load pattern of weekdays 6 indicated by the solid line is not similar to the load patterns of weekdays 1 to 5, but similar to the load patterns of holidays 1 to 3. The weekday 6 is described as a weekday in FIG. 4 because the day type of the target day recorded in the calendar information recording unit 11 is a weekday.

In such a case, the day type determination unit 104 determines the similarity between the load pattern on weekday 6 and the load patterns on holiday 1 to holiday 3 SI holiday _log ￣ (weekday 6) (actual ￣ is “SI holiday _log ” Is determined to satisfy the condition shown in the following (Equation 7).

Here, the day type determination unit 104 determines that the similarity SI holiday _log休 日 (weekday 6) based on the load pattern of weekday 6 shown in FIG. 4 satisfies the condition shown in (Equation 7). That is, the day type determination unit 104 determines the day type of the load pattern of weekday 6 as a holiday. As a result, the day type determination unit 104 corrects the day type of the target day from weekdays to holidays, and outputs the result to the determination result recording unit 105.

Next, the correction processing of the day type (correction processing from weekday to unique day) in the day type determination unit 104 will be described using a specific example 2 of the load pattern.
FIG. 5 is a diagram illustrating a specific example 2 of the load pattern for explaining the correction processing of the day type in the day type determination unit 104. As shown in FIG. 5, the load patterns on weekdays 1 to 5 are similar among the respective load patterns. The load patterns of holidays 1 to 4 are also similar among the load patterns. However, the load patterns on weekdays 6 indicated by solid lines are different in the following points when the load patterns on weekdays 1 to 5 are compared. The load pattern on weekday 6 is different in that the maximum value of the load amount is about 2000, and the load pattern on weekdays 1 to 5 is that the maximum value of the load amount is included in the range of 2500 to 3200. The load pattern on weekday 6 starts falling after about 12:00, but the load pattern on weekdays 1 to 5 differs in that the load starts dropping from about 16:00.

  Also, comparing the load pattern on weekday 6 with the load patterns on holidays 1 to 4, the load pattern on weekday 6 has a peak exceeding a load amount of 1000, but the load pattern on holidays 1 to holiday 4 is almost flat. In that there is no peak exceeding a load of 1000. That is, the load pattern on weekdays 6 is not similar to the load patterns on weekdays 1 to 5, and is not similar to the load patterns on holidays 1 to 3. The weekday 6 is, for example, the calendar information recording unit 11 in which the day type of the target day is erroneously set to a weekday.

In such a case, the day type determination unit 104 determines the similarity between the load pattern of weekday 6 and the load patterns of holiday 1 to holiday 4 SI holiday _log ￣ (weekday 6) by the condition shown in the above-described (formula 7). It is determined whether or not the condition is satisfied.

Here, the day type determination unit 104 determines that the similarity SI holiday _log ￣ (weekday 6) based on the load pattern of weekday 6 shown in FIG. 5 does not satisfy the condition shown in (Equation 7). Next, the day type determination unit 104 determines the similarity between the load pattern of weekday 6 and the load patterns of weekday 1 to weekday SI weekday _log ￣ (weekday 6) (actual ￣ is above “SI weekday _log ”). Is determined as to whether or not the following condition (Expression 8) is satisfied.

Here, the day type discrimination unit 104, similarity SI weekday _{log ¯} based on load patterns of weekdays 6 shown in FIG. 5 _(weekdays 6) is determined not to satisfy the conditions shown in (Equation 8). That is, the day type determination unit 104 determines that the day type of the load pattern of weekday 6 is neither a weekday nor a holiday, and thus is a unique day. As a result, the day type determining unit 104 corrects the day type of the target day from weekdays to special days, and outputs the corrected day type to the determination result recording unit 105.

Next, a description will be given of an operation of the load prediction device 10 for discriminating the day type by using the above-described second discrimination method of discriminating the day type using the intermediate value.
FIG. 6 is a flowchart showing an operation relating to discrimination of a day type in the load prediction device 10. As shown in FIG. 6, in step S100, the load result data recording unit 101 receives the load result data of each time zone from the load result collection unit 13, records the data in association with a date, and sequentially stores the data.

  Next, in step S110, the similarity calculating unit 102 calculates the similarity of the load pattern between the target date and each day of the target period based on the load actual data referred from the load actual data recording unit 101, The similarity table 103 is updated. Next, in step S120, the day type determination unit 104 refers to the similarity in the day similarity table 103 and determines the day type of the target day using the above-described second determination method. Next, in step S130, the day type determination unit 104 refers to the day type of the target date from the calendar information recording unit 11 and determines whether the date type is different from the day type of the determination result (= whether the date type is corrected or not). Judge).

  Here, when it is determined that the day type is to be corrected (YES in step S130), the process proceeds to step S140, and the day type determining unit 104 corrects the day type referred to from the calendar information recording unit 11 to the day type of the determination result. I do. Next, proceeding to step S150, the day type determining unit 104 outputs the corrected day type to the determination result recording unit 105.

  If it is determined that the day type is not to be corrected (NO in step S130), the process proceeds to step S150, and the day type determining unit 104 outputs the uncorrected day type to the determination result recording unit 105. Through the above processing, the determination result recording unit 105 can accumulate the corrected date type in association with the target date.

  By the operation described above, the load prediction device 10 determines the day type of the target day, so that even if the day type referred to by the calendar information recording unit 11 is incorrect, the determination result recording unit 105 The correct type of day after the correction can be recorded.

Next, a process in which the abnormality detection unit 108 of the load prediction device 10 detects an abnormal value will be described using a specific example 3 of the load pattern.
FIG. 7 is a diagram illustrating a specific example 3 of the load pattern for describing a process in which the abnormality detection unit 108 detects an abnormal value. As shown in FIG. 7, the load patterns of holidays 1 to 4 are similar among the load patterns. However, the load pattern of the holiday 5 indicated by the solid line has an abnormal value in the time zone of 21:00 to 23:00. The abnormality detection unit 108 detects, as an abnormal value, the load amount after 21:00 in the load pattern of the holiday 5 based on the statistical value corresponding to the day type of the target day, referred to from the day type information 106.

Here, a specific example of a statistical value will be described, and a specific abnormal value detection process will be described. The day type determining unit 104 determines, as one of the statistical values, the average value W _I (t) ￣ (actual) of each time period t in the load pattern of the day type I (for example, I = weekday) included in the target period S. Is located above “W _I (t)” using the following (Equation 9). Note that, in (Equation 9), i = 1 to n indicate the corresponding day in the target period S of the day type I. In addition, the day type determination unit 104 calculates the standard deviation assumed value SSD _log (t) ￣ (actual ￣ is located above “SSD _log (t)”) as one of the above statistic values. It is calculated using the following (Equation 10).

The statistical values calculated by the day type determination unit 104 are stored in the day type information 106. Next, the abnormality detection unit 108 obtains the average value W _l (t) ￣ (the actual ￣ is located above “W ₁ (t)”) and the standard deviation assumed value SSD _log ( t)}, an abnormality of the load W (t) in the time zone t is detected by the following (Equation 11). In (Equation 11), the average value W _l (t) ￣ is the average value of the time period t of the load pattern having the same day type 1 as the load amount W (t) to be detected. As shown in (Equation 11), the abnormality detection unit 108 calculates the absolute value of the difference between the load amount W (t) to be detected and the average value W ₁ (t) ￣ as the average value W ₁ (t) ￣. Is larger than 3σ (SSD _log (t) ￣), it is detected as an abnormal value. Also, t = 1,..., M (m = 48: when the unit of the time zone is 30 minutes).

Next, a process in which the abnormality detection unit 108 of the load prediction device 10 detects an abnormal value will be described using a specific example 4 of a load pattern having large variations.
FIG. 8 is a diagram illustrating a specific example 4 of the load pattern for describing a process in which the abnormality detection unit 108 detects an abnormal value. As shown in FIG. 8, the variation in each time zone is large until about 11:00, and the variation is small after 14:00. If there are few load patterns of the same day type, the variation for each time zone may be large.

In the case of such a load pattern, the day type determination unit 104 obtains the expected standard deviation SSD _log (t) using the following (Equation 12) instead of the above (Equation 10). In (Equation 12), CV￣ (actual ￣ is located above “CV”) is the average value of the variation coefficient CV (t), and the day type determination unit 104 calculates the following (Equation 13) Is determined using Note that t = 1,..., M (m = 48: when the unit of the time zone is 30 minutes). In the equation (13), CV (t) is a variation coefficient in the time zone t, and the day type determination unit 104 calculates CV (t) for the entire time zone using the following (Equation 14).

  As described above, when the variation in the load pattern is large, the abnormality detection unit 108 of the load prediction device 10 can accurately detect the abnormal value using (Equation 12).

Next, a process in which the abnormality detection unit 108 of the load prediction device 10 corrects an abnormal value of the load pattern will be described. The abnormality detection unit 108 calculates the average of the load in the time zone other than the abnormal value and the load in the time zone other than the abnormal value using the following (Equation 15) for the load amount in the time zone in which the abnormal value is detected. And the average value of the deviation is corrected by a value added to the average value W _l (t) ￣.

  FIG. 9 is a diagram illustrating a correction example in which the abnormality detection unit 108 of the load prediction device 10 corrects an abnormal value of a load pattern. FIG. 9 shows the measured holiday load patterns and the average, upper limit, and lower limit of the load patterns obtained as statistical values. As shown in FIG. 9, the holiday load pattern after 20:00 indicates an abnormal value exceeding the upper limit load pattern. By correcting the abnormal value by the abnormality detecting unit 108, a load pattern falling within the upper limit load pattern is obtained as shown in the correction part 90 surrounded by the dashed line ellipse.

(Second embodiment)
Next, as a second embodiment, a load prediction device having a function of monitoring a change in actual load data on the current day and determining a day type will be described.
FIG. 10 is a diagram illustrating a configuration example of a load prediction device according to the second embodiment. As illustrated in FIG. 10, the load prediction device 10a differs from the load prediction device 10 illustrated in FIG. 1 in that the load prediction device 10a includes a prediction load pattern extraction unit 107a and a deviation detection unit 114. Is the same. Therefore, the same components are denoted by the same reference numerals as those in FIG. 1 and the description is omitted. Although not shown in FIG. 10, the operation system including the load prediction device 10a has the same configuration as the operation system 1 shown in FIG.

  The deviation detection unit 114 refers to the statistical value of each day from the day type information 106, refers to the day type of the current day (the day during which actual load data is being collected) from the calendar information recording unit 11, and 101, the load pattern of the current day is referred to, and based on the statistical value corresponding to the day type of the current day, the normal range in which the load pattern is predicted for each time period up to the present time from the statistical value of the corresponding time period. This is detected when it deviates from. Further, when a deviation of the load pattern is detected, the deviation detection unit 114 has a function of estimating a correct day type of the day and outputting a correction instruction of the day type to the prediction load pattern extraction unit 107a. The prediction load pattern extraction unit 107a outputs the load pattern of the target period for each day according to the correction instruction from the deviation detection unit 114 as the prediction load pattern.

Specifically, the divergence detection unit 114 calculates the average value W _I (t) of the day type I set in the time zone t as the statistical value of the day type from the day type information 106 and the standard deviation assumed value SSD. _With reference to _log (t) ￣, a deviation is detected when the following condition (Equation 16) is satisfied. In (Equation 16), the day type of the day referred to from the calendar information recording unit 11 is the day type I. In (Equation 16), t is a time zone immediately before the current time.

When the divergence of the load pattern is detected by the above processing, the divergence detection unit 114 determines that there is a day type J that continuously satisfies the following condition (Equation 17) over a predetermined number of time periods. An instruction to correct the type is transmitted to the prediction load pattern extraction unit 107a. Thereby, the prediction load pattern recording unit 110 stores the prediction load pattern corresponding to the correct day type. The load pattern prediction unit 112 can output a load prediction pattern based on a load pattern for prediction corresponding to a correct day type. The device planning unit 15 can generate an operation pattern based on a load prediction pattern corresponding to a correct day type. As described above, the operation system including the load prediction device 10a can control the energy device to be managed based on the operation pattern corresponding to the correct day type J, although it is from the middle of the day.

  FIG. 11 is a diagram illustrating an example of a deviation in the load pattern. FIG. 11 shows the load patterns on weekdays during actual measurement and the average, upper limit, and lower limit of the load patterns on weekdays and holidays obtained as statistical values. The weekdays during actual measurement shown in FIG. 11 are, for example, those in which the calendar information recording unit 11 incorrectly sets the day type of the target day to weekdays. As shown in FIG. 11, the day type falls below the weekday lower limit load pattern after about 6:00 on the current day, and a deviation occurs. Then, as shown in FIG. 11, since several consecutive hours are included in the range from the upper limit load pattern to the lower limit load pattern, the day type is corrected from weekdays to holidays. In this way, the divergence detection unit 114 detects the divergence as shown in FIG. 11 by processing the load pattern of the weekday during the actual measurement for each time zone, and corrects the day type of the load pattern after the divergence to a holiday. I do.

  In each of the above embodiments, the similarity calculation unit 102, the day type determination unit 104, the prediction load pattern extraction unit 107, the abnormality detection unit 108, the change notification unit 109, the load pattern prediction unit 112, the load prediction The prediction load pattern extraction unit 107a and the deviation detection unit 114 in the device 10a are assumed to be software function units, but may be hardware function units such as LSIs.

  According to at least one embodiment described above, the similarity calculating unit 102 that calculates the similarity of the load pattern for each day in the predetermined period S, and the load based on the similarity calculated by the similarity calculating unit 102. By having the day type discriminating unit 104 for discriminating the day type of the pattern, it is possible to exclude disturbance data from the actual load data.

  Further, the similarity calculation unit 102, the day type determination unit 104, the prediction load pattern extraction unit 107, the abnormality detection unit 108, the change notification unit 109, the load pattern prediction unit 112, the load prediction When the functions of the prediction load pattern extraction unit 107a and the deviation detection unit 114 in the device 10a are realized by software, a program for realizing those functions is recorded on a computer-readable recording medium, and The program may be read by a computer system and executed. Here, the “computer system” includes an OS (Operating System) and hardware such as peripheral devices. The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD (Compact Disk) -ROM, and a storage device such as a hard disk built in a computer system. . Further, the “computer-readable recording medium” refers to a volatile memory (RAM) inside a computer system which is a server or a client when a program is transmitted through a network such as the Internet or a communication line such as a telephone line. In addition, programs that hold programs for a certain period of time are also included.

Further, the above program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting a program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
Further, the above program may be a program for realizing a part of the functions described above. Furthermore, the above-mentioned program may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system. Further, the operation system 1 may be constructed using cloud computing.

  While some embodiments of the invention have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. These embodiments can be implemented in other various forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are also included in the invention described in the claims and equivalents thereof.

DESCRIPTION OF SYMBOLS 1 ... Operation system, 2 ... Load equipment, 4 ... Energy equipment, 10 and 10a ... Load prediction device, 11 ... Calendar information recording part, 12 ... Input / output part, 13 ... Load result collection part, 15 ... Equipment planning part, 16 ... Operation pattern recording unit, 17 ... Device control unit, 101 ... Load performance data recording unit, 102 ... Similarity calculation unit, 103 ... Day similarity table, 104 ... Day type discrimination unit, 105 ... Discrimination result recording unit, 106 ... Day type information, 107, 107a: prediction load pattern extraction unit, 108: abnormality detection unit, 109: change notification unit, 110: prediction load pattern recording unit, 111: weather information recording unit, 112: load pattern prediction unit, 113: load prediction pattern recording unit; 114: deviation detection unit

Claims (10)

A similarity between a plurality of load patterns is obtained based on a load pattern indicating a transition of a load in a predetermined period, which is obtained based on an actual power amount to be managed and a contract power amount in the predetermined time period at predetermined time intervals. A similarity acquisition unit that performs
Based on the similarity acquired by the similarity acquisition unit, a type determination unit that determines the type of the predetermined period in which the load pattern is obtained,
A load performance data discriminating device comprising: The type determination unit corrects a preset type to a type determined by the type determination unit when the type determined by the type determination unit differs from a type determined in advance for each of the predetermined periods. Load actual data discriminator. A similarity obtaining unit that obtains a similarity between a plurality of load patterns based on a load pattern indicating a transition of a load for a predetermined period obtained by measuring a load to be managed for each predetermined time;
Based on the similarity acquired by the similarity acquisition unit, a type determination unit that determines the type of the predetermined period in which the load pattern is obtained,
A determination result recording unit that records the type determination result in the type determination unit in association with information that specifies the predetermined period,
An extraction unit configured to extract a plurality of load patterns used to predict the load in association with a type of the determination result, based on the determination result recorded in the determination result recording unit;
Equipped with a,
The type determination unit acquires a statistical value for each type from a plurality of the load patterns,
A statistical value storage unit that stores the statistical value obtained by the type determination unit,
An abnormality detection unit that detects an abnormal value of the load pattern based on the statistical value obtained from the statistical value storage unit;
Further comprising
The abnormality detection unit corrects the detected abnormal value based on a deviation between the load of the predetermined period other than the abnormal value and the average of the load of the predetermined period other than the abnormal value,
Load prediction device. A similarity between a plurality of load patterns is obtained based on a load pattern indicating a transition of a load in a predetermined period, which is obtained based on an actual power amount to be managed and a contract power amount in the predetermined time period at predetermined time intervals. A similarity acquisition unit that performs
Based on the similarity acquired by the similarity acquisition unit, a type determination unit that determines the type of the predetermined period in which the load pattern is obtained,
A determination result recording unit that records the type determination result in the type determination unit in association with information that specifies the predetermined period,
An extraction unit configured to extract a plurality of load patterns used to predict the load in association with a type of the determination result, based on the determination result recorded in the determination result recording unit;
A load prediction device comprising: A notification unit that notifies a user when the type referred to from the determination result recording unit is different from the type set in advance for each of the predetermined periods,
The load prediction device according to claim 3, further comprising: From the statistic storage unit, a statistic corresponding to the type of the predetermined period and corresponding to the immediately preceding time zone is obtained, and the load pattern up to the immediately preceding time zone is predicted from the statistic value. The load prediction device according to claim 3 , further comprising a deviation detection unit that detects deviation from a range to be performed. The load prediction device according to claim 6 , wherein when detecting the deviation, the deviation detection unit estimates a correct type based on a load pattern up to the immediately preceding time zone. A similarity between a plurality of load patterns is obtained based on a load pattern indicating a transition of a load in a predetermined period, which is obtained based on an actual power amount to be managed and a contract power amount in the predetermined time period at predetermined time intervals. Similarity acquisition step to be performed,
A type determining step of determining a type of the predetermined period in which the load pattern is obtained based on the similarity obtained in the similarity obtaining step;
Load data determination method having the following. A similarity obtaining step of obtaining a similarity between a plurality of load patterns based on a load pattern indicating a transition of a load for a predetermined period obtained by measuring a load of a management target for each predetermined time;
A type determining step of determining a type of the predetermined period in which the load pattern is obtained based on the similarity obtained in the similarity obtaining step;
A determination result recording step of recording the determination result of the type in the type determination step in a determination result recording unit in association with information specifying the predetermined period;
An extraction step of extracting a plurality of load patterns used for predicting the load in association with a type of the determination result, based on the determination result recorded in the determination result recording unit,
Have a,
Obtaining a statistical value for each type from a plurality of the load patterns in the type determination step,
A statistic value storing step of storing the statistic value obtained in the type determination step,
An abnormality detecting step of detecting an abnormal value of the load pattern based on the statistical value obtained in the statistical value storing step;
Further having
In the abnormality detection step, based on a deviation between the load of the predetermined period other than the abnormal value and the average of the load of the predetermined period other than the abnormal value, correct the detected abnormal value,
Load prediction method. A similarity between a plurality of load patterns is obtained based on a load pattern indicating a transition of a load in a predetermined period, which is obtained based on an actual power amount to be managed and a contract power amount in the predetermined time period at predetermined time intervals. Similarity acquisition step to be performed,
A type determining step of determining a type of the predetermined period in which the load pattern is obtained based on the similarity obtained in the similarity obtaining step;
A determination result recording step of recording the determination result of the type in the type determination step in a determination result recording unit in association with information specifying the predetermined period;
An extraction step of extracting a plurality of load patterns used for predicting the load in association with a type of the determination result, based on the determination result recorded in the determination result recording unit,
Load prediction method having the following.

Images