JP2018124727A - Electric power demand prediction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve accuracy of prediction of an electronic power demand in the future considering an event.SOLUTION: In an electric power demand prediction device according to an embodiment, an event determination section extracts event time series data that has influenced on an electric power demand from event time series data having an influence degree of an event indicating an influence degree of the event to the electric power demand for a prescribed unit time on the basis of an amount of a demand fluctuation that is a difference between a predicted demand calculated without considering the influence of the event and an actual demand that is an electric power demand actually occurred in a time zone indicated by the predicted demand, and saves the data as a past event time series data with mapping the event time series data with the amount of the demand fluctuation. A demand fluctuation prediction section predicts a real time amount of the demand fluctuation by using an amount of a demand fluctuation of a similar past event time series data and an influence degree of the event, and calculates a result of demand prediction considering the event by adding a normal demand to the amount of the demand fluctuation.SELECTED DRAWING: Figure 2

Description

  Embodiments described herein relate generally to a power demand prediction apparatus.

  Since the retailing of electricity was fully liberalized in April 2016, many new power companies have entered the electricity business. For a new power company, it is important to predict future power demand with high accuracy, for example, in units of 30 minutes.

  Conventionally, there is a technique for predicting power demand based on information on weather conditions (temperature, humidity, precipitation, etc.). Electricity demand is also affected by events (sports competitions, concerts, fireworks festivals, Obon, New Year, disasters, etc.) in addition to weather conditions. Therefore, as a conventional technique, for example, there is one that predicts power demand based on event data input by each consumer.

JP 2006-157984 A

  However, further improvement is desired regarding the accuracy of prediction of future power demand in consideration of events.

  In the power demand prediction apparatus of the embodiment, the event determination unit does not consider the influence of the event from the event time-series data having the event influence degree indicating the influence degree of the event with respect to the power demand every predetermined unit time. Event time-series data that affected power demand based on the demand fluctuation amount, which is the difference between the forecast demand calculated in step 1 and the actual demand that was actually generated in the time zone indicated by the forecast demand. And the event time-series data and the demand fluctuation amount are associated with each other and stored as past event time-series data. The normal demand prediction unit predicts normal demand for each prediction target area without considering the influence of the event. The similar event determination unit acquires or generates real-time event time-series data having an event influence degree indicating the influence degree of the real-time event with respect to power demand for each predetermined unit time, and generates a plurality of past event time series. It is determined from the data whether there is something similar to the real-time event time-series data. The variable demand forecasting unit predicts the real-time demand fluctuation amount using the demand fluctuation amount and event influence degree of similar past event time series data, and adds the normal demand to the demand fluctuation amount, thereby Calculate the demand forecast result in consideration.

FIG. 1 is a diagram illustrating a hardware configuration of the power demand prediction apparatus according to the first embodiment. FIG. 2 is a diagram illustrating a functional configuration of the power demand prediction apparatus according to the first embodiment. FIG. 3 is a diagram illustrating a functional configuration of a data organizing unit of the power demand prediction apparatus according to the first embodiment. FIG. 4 is a diagram illustrating an event table according to the first embodiment. FIG. 5 is a diagram illustrating an event table with an event influence degree according to the first embodiment. FIG. 6 is a diagram illustrating an event similarity table according to the first embodiment. FIG. 7 is a diagram illustrating event time-series data according to the first embodiment. FIG. 8 is a diagram illustrating a functional configuration of an event determination unit of the power demand prediction apparatus according to the first embodiment. FIG. 9 is a graph showing the change over time in the demand fluctuation amount in the first embodiment. FIG. 10 is a graph showing the change over time in the demand fluctuation amount in the first embodiment. FIG. 11 is a diagram illustrating a functional configuration of a demand prediction unit of the power demand prediction apparatus according to the first embodiment. FIG. 12 is a graph showing a change over time in the real-time demand fluctuation amount in the first embodiment. FIG. 13A is a diagram showing past event time-series data in the first embodiment. FIG. 13B is a diagram showing real-time event time-series data in the first embodiment. FIG. 14 is a flowchart illustrating power demand prediction processing by the power demand prediction apparatus according to the first embodiment. FIG. 15 is a diagram illustrating an event table according to the second embodiment. FIG. 16 is a diagram illustrating a functional configuration of an event determination unit of the power demand prediction apparatus according to the second embodiment. FIG. 17 is a diagram illustrating a functional configuration of a power demand prediction apparatus according to a comparative example.

  Hereinafter, a first embodiment and a second embodiment of a power demand prediction apparatus of the present invention will be described with reference to the drawings. In these embodiments, as an example, a case where a future power demand is predicted in units of 30 minutes after one hour will be described. In the following, power demand is also simply referred to as “demand”. In addition, the real-time may include not only the most recent few minutes but also the most recent hours or days.

  An event refers to an event or event that affects the power demand. Specifically, for example, sports events, concerts, fireworks events, New Year homecoming, Obon homecoming, disasters, road congestion, epidemic diseases, and the like. The event data is information regarding events. For example, event data of a sports tournament is information such as the date and time of the event (occurrence), location, and contents. In addition, for example, the event data for the New Year's homecoming includes information on the reservation status of airplanes, bullet trains, hotels, restaurants, etc. during the New Year period. The event data is information that can be obtained from SNS (Social Network Service) on TV, radio, and the Internet, various websites, and the like.

  In addition, future power demand is predicted for each area, not for each consumer. Here, the area is an area including a plurality of consumers as a target when collecting past demand data or predicting future demand. Specifically, for example, the area may be a prefecture unit, a municipal unit, or a predetermined unit separated by a distance from the place where the event is held, but is not limited thereto. Further, the customers in the area are not limited to all customers, and may be only customers who have contracted with a specific new electric power company. Note that an area for which future demand is predicted is also referred to as a “forecast target area”.

  Moreover, when collecting the past demand data for every area, the following methods can be considered. For example, if at least one of smart meters (devices that digitally measure power consumption and send measurement results to the outside) and HEMS (Home Energy Management System) is installed in all of the target customers in the area, What is necessary is just to total as demand data every 30 minutes based on the data from them. In addition, if there is a customer who has neither smart meter nor HEMS installed, the demand of the entire area is estimated by statistical estimation from the data of the customer where they are installed, and the demand data every 30 minutes It is good.

Here, in order to help understanding, first, a power demand prediction apparatus of a comparative example (prior art) will be described. FIG. 17 is a diagram illustrating a functional configuration of a power demand prediction apparatus according to a comparative example. For example, the demand prediction unit of the power demand prediction apparatus of the comparative example performs demand prediction in consideration of an event for each area by the following equation (1).
y _(t) = Σα _{i (t)} x _{i (t)} + C (1)

Here, t is an identifier representing time. y _(t) is the predicted demand. x _{i (t)} is the weather condition (temperature, humidity, precipitation, etc.). α _{i (t)} is a coefficient, and can be calculated in advance by, for example, the least square method based on the relationship between past weather conditions and demand. C is a constant based on the amount of change in demand that has occurred during similar events in the past.

  According to this equation (1), a demand prediction result can be obtained for each area using the demand data, weather data, and event data for each area (area A, area B,...). That is, it is possible to perform demand prediction in consideration of events. However, since C is a constant, it is impossible to consider the influence of the event on the demand for every 30 minutes in the same event. In addition, there are cases where it is not possible to accurately determine which past event affects the demand to the same extent as the influence of the event to be considered from now on. For example, when trying to decide by event name, there may actually be cases where the event name is the same but the impact on demand is different, or even if the event name is different, the impact on demand is the same. It is. Therefore, the power demand prediction apparatus of this embodiment was created to improve these points.

(First embodiment)
Hereinafter, the power demand prediction apparatus of the first embodiment will be described. Note that in the first embodiment, an event having the same degree of influence on demand is assumed as an event across several areas. As such events, a predetermined international sports tournament (hereinafter simply referred to as “international sports tournament”) and homecoming in Obon (hereinafter simply referred to as “homecoming”) will be described as examples.

  FIG. 1 is a diagram illustrating a hardware configuration of a power demand prediction apparatus 1 according to the first embodiment. The power demand prediction device 1 is a computer device that predicts future power demand for each area. The power demand prediction apparatus 1 is, for example, a personal computer (PC), which is a CPU (Central Processing Unit) 2, a memory 3 (storage means), a hard disk 4 (storage means), a GUI (Graphical User Interface) connected to each other via a bus 7. ) 5 and an interface 6.

  A program for realizing each function described later in the power demand prediction apparatus 1 is stored, for example, on the hard disk 4, expanded on the memory 3 at the time of execution, and then executed by the CPU 2 according to the procedure. Event data, various area demand data, and weather data are input to the interface 6. In addition, the interface 6 outputs various area demand prediction results.

  Each data (intermediate data) generated in the middle of the process may be managed on the memory 3 or stored on the hard disk 4. Each of these data may be visualized by the operator through the GUI 5.

  FIG. 2 is a diagram illustrating a functional configuration of the power demand prediction apparatus 1 according to the first embodiment. As illustrated in FIG. 2, the power demand prediction apparatus 1 includes a data organization unit 11, an event determination unit 12, and a demand prediction unit 13 as functional configurations.

  As described above, event data, various area demand data (area A demand data, area B demand data), and weather data (area A weather data, area B weather data) are input to the power demand prediction apparatus 1. In addition, the power demand prediction apparatus 1 calculates various area demand prediction results (area A demand prediction result, area B demand prediction result). Moreover, the power demand prediction apparatus 1 generates and holds event time-series data, various area demand fluctuation data (area A demand fluctuation data, area B demand fluctuation data), and the like as intermediate data.

  Hereinafter, a procedure for calculating various area demand prediction results will be described with a description of data and a description of processing. The weather data is data obtained by collecting weather condition information for each area. The meteorological data is composed of, for example, area, date / time, temperature, humidity, and precipitation information. The weather data is fact data for past data and forecast data for future data.

  Various area demand data are totaled as demand data every 30 minutes based on the data of a smart meter and HEMS, for example as mentioned above. The event data is acquired from, for example, an SNS or various websites on the television, radio, and the Internet as described above (details will be described later).

Here, an outline of processing by the power demand prediction apparatus 1 will be described. The power demand prediction apparatus 1 calculates the predicted demand by the following equation (2).
y _(t) = Σα _{i (t)} x _{i (t)} + Σβ _{i (t)} E _{i (t)} Expression (2)

Here, t is an identifier representing time. y _(t) is the predicted demand. x _{i (t)} is the weather condition (temperature, humidity, precipitation, etc.). α _{i (t)} is a coefficient, and can be calculated in advance by, for example, the least square method based on the relationship between past weather conditions and demand. E _{i (t)} is an event influence degree (details will be described later). β _{i (t)} is a coefficient, and can be calculated in advance by, for example, the least square method based on the relationship between past events and demand.

  By the processing using the formula (2), for example, the influence on the demand due to the event can be taken into consideration for each time period of 30 minutes within the same event. In addition, it is possible to accurately determine past events that affect the demand to the same extent as the influence of the event to be considered from now on. For example, past events with the same event name but different impact on demand are not used for processing, or past events with the same impact on demand even with different event names are used for processing. can do.

  Hereinafter, the configuration and processing details of the power demand prediction apparatus 1 will be described. FIG. 3 is a diagram illustrating a functional configuration of the data organizing unit 11 of the power demand prediction apparatus 1 according to the first embodiment. As shown in FIG. 3, the data organizing unit 11 includes an event table generating unit 111, an event influence degree calculating unit 112, and a similarity degree judging unit 113.

  The event table generation unit 111 inputs event data that is data related to an event, and generates an event table having at least an event name and an event property including an event occurrence time and an end time. FIG. 4 is a diagram illustrating an event table according to the first embodiment. The event-related properties include, for example, event occurrence time, event end time, information distribution time, hash tag (#xxx), keyword, information sender, and the like.

  Returning to FIG. 3, the event influence degree calculation unit 112 calculates an event influence degree (details will be described later) based on information acquired from one or more of television, radio, and the Internet regarding the event, and an event table An event table with an event impact level is generated by adding the event impact level in association with the event.

  FIG. 5 is a diagram illustrating an event table with an event influence degree according to the first embodiment. Here, the event influence degree indicates the influence degree of the event on the power demand. More specifically, the event influence degree is an index indicating the number of people who are affected by the occurrence of an event. The event influence degree is calculated from, for example, the number of postings of event data, the influence degree of the poster, the reliability of the poster, and the like. Here, the influence level of the poster is calculated from, for example, the population of the broadcast area, the audience rating, and the like in the case of television and radio, and is calculated from the number of followers in the case of SNS. Further, the reliability of the poster is calculated from the credibility of information provided in the past by the poster. It should be noted that a constant value of event influence may be used during the event.

  Returning to FIG. 3, the similarity determination unit 113 determines the similarity between events based on one or more of the event name and the property in the event table with event influence degree, and a plurality of similar events. Event time-series data having an event influence degree for each predetermined unit time is generated based on the event influence degree associated with each of the above.

  Note that the similarity is calculated, for example, as the number of event names and properties that the event has that includes the same name and time. It is also possible to weight each element.

  The similarity determination unit 113 determines the event similarity using, for example, an event similarity table. FIG. 6 is a diagram illustrating an event similarity table according to the first embodiment. The event similarity table in FIG. 6 is created using the event table with event influence degree in FIG. Here, the event name is given twice the weight. FIG. 7 is a diagram illustrating event time-series data according to the first embodiment.

  For example, in FIG. 5, event A and event C have the same event name “International Sports Tournament”, and the event name is weighted twice, so the similarity is “2”. ing. Event A and event C are combined to create event time-series data of the international sports event shown in FIG.

  In FIG. 5, event B and event D are common in that the event name includes “homecoming”, and the event name is doubled in weight, so the degree of similarity based on it Is “2”. In addition, since the property “hashtag, keyword” is the same for “Obon”, the similarity is “1”. Therefore, the similarity between the event B and the event D is “3” in total. Event B and event D are combined to generate homecoming event time-series data shown in FIG. 7B.

  Thus, based on the event name and the properties of the event, the similarity of the event can be determined easily and with high accuracy. Here, only the common contents are used for determining the similarity, but if the contents are not common contents but have high correlation, they may be used for determining the similarity according to the level of the correlation. Good. For example, “homecoming” and “homecoming” may have a predetermined similarity.

  Next, the event determination unit 12 shown in FIG. 2 will be described. FIG. 8 is a diagram illustrating a functional configuration of the event determination unit 12 of the power demand prediction apparatus 1 according to the first embodiment. As shown in FIG. 8, the event determination unit 12 includes a demand fluctuation determination unit 121 and an event data extraction unit 122.

  The demand fluctuation determination unit 121 inputs demand fluctuation data for each prediction target area, and determines the time (time zone) when the demand fluctuation occurs and the demand fluctuation amount. Here, the demand fluctuation data for each prediction target area is obtained by calculating the difference between the predicted demand (for example, calculated by the conventional method of FIG. 17) and the actual demand for each prediction target area in units of 30 minutes. The presence or absence of demand fluctuation is determined by, for example, a threshold value. As the threshold value, for example, a constant of demand fluctuation amount, a rate of change of demand fluctuation with time, a rate of increase from the average demand amount of each area, and the like are used.

  Here, FIG. 9 is a graph showing the change over time in the demand fluctuation amount in the first embodiment. In FIG. 9, a constant of demand fluctuation amount is used as the threshold value. Each plotted point shows the demand fluctuation amount. Then, it can be seen that the time zone in which the demand fluctuates, that is, the time zone in which the threshold value is exceeded is 23: 00 to 2:30.

  Returning to FIG. 8, the event data extraction unit 122 uses the predicted demand calculated from the event time series data without considering the influence of the event and the power demand actually generated in the time zone indicated by the predicted demand. Based on the demand fluctuation amount that is the difference from a certain actual demand, the event time series data that affected the power demand is extracted, the event time series data and the demand fluctuation amount are associated with each other and the past event time series data. Save as. More specifically, the event data extraction unit 122 is, for example, about a time zone in which it is determined that there is a demand fluctuation based on the demand fluctuation amount from the event time series data (may include a time zone before and after that). Then, the event time series data of the event that has affected the power demand is extracted, and the event time series data and the demand fluctuation amount are associated with each other and stored as past event time series data.

  For example, the event data extraction unit 122 may measure the similarity (duplication degree) between the time zone in which the demand fluctuation is determined and the occurrence time zone of the event, and the fluctuation in the event influence degree of the demand fluctuation waveform and the event time-series data. The event time series data of the event that has influenced the power demand is extracted based on at least one of the similarity to the waveform. For example, in the example shown in FIG. 10, when there are two types of events, an international sports event and a homecoming event, the similarity (for example, the least square method) between the demand fluctuation waveform and the event influence degree fluctuation waveform of the event time-series data The international sports tournament having a high calculation) is extracted as an event that has influenced the power demand and stored as past event time series data. In this way, the event determination unit 12 can accurately extract event time series data of events that affect demand fluctuation using the event time series data and the demand fluctuation data for each prediction target area.

  Next, the demand prediction unit 13 shown in FIG. 2 will be described. FIG. 11 is a diagram illustrating a functional configuration of the demand prediction unit 13 of the power demand prediction apparatus 1 according to the first embodiment. As shown in FIG. 11, the demand prediction unit 13 includes a normal demand prediction unit 131, a similar event determination unit 132, and a variable demand prediction unit 133.

  The normal demand prediction unit 131 predicts the normal demand for each prediction target area without considering the influence of the event. More specifically, the normal demand prediction unit 131 predicts normal demand for each prediction target area from weather data. Here, the normal demand is a demand amount calculated from a demand prediction model using conventional weather data as an explanatory variable, and does not consider event data.

  The similar event determination unit 132 acquires or generates real-time event time-series data having an event influence degree indicating the influence degree of the real-time event on the power demand for each predetermined unit time (similar to the data organizing part 11). It is determined whether there is something similar to the real-time event time-series data from a plurality of past event time-series data. For example, the similar event determination unit 132 determines whether there is a thing similar to real-time event time-series data from a plurality of past event time-series data, a time zone in which it is determined that there is a demand fluctuation, and an event occurrence time zone And the similarity between the waveform of the demand fluctuation and the waveform of the event influence degree fluctuation of the event time-series data. This process is started, for example, when a demand fluctuation exceeding the threshold value occurs in the prediction target area or when an event having an event influence level exceeding the threshold value occurs.

  FIG. 12 is a graph showing a change over time in the real-time demand fluctuation amount in the first embodiment. Each plotted point shows real-time demand fluctuation. In the example of FIG. 12, the demand fluctuation amount exceeds the threshold at 23:00, and the process by the similar event determination unit 132 is started. In that case, the similar event determination unit 132 refers to the real-time event time-series data for a time after a time (for example, one hour before) slightly before the time when the demand fluctuation amount exceeds the threshold, and affects the power demand. Extract the given event time series data. To extract the event time-series data that has affected, for example, the degree of similarity (duplication) between the time zone where the demand fluctuation occurs and the time when the event is held (occurrence), the waveform of the demand fluctuation and the event time series The similarity with the waveform of the fluctuation of the event influence degree of data is used.

  When the similar event determination unit 132 extracts real-time event time-series data that affects demand fluctuation, the similar event determination unit 132 searches for similar events from past event time-series data. The presence / absence of this similar event is determined based on the event name, the property of the event, and the event influence level.

  Here, FIG. 13A is a diagram illustrating past event time-series data in the first embodiment. FIG. 13B is a diagram showing real-time event time-series data in the first embodiment. When the real-time event time-series data shown in FIG. 13B is present, the event name matches with “International Sports Tournament” and the manner of increasing the event influence degree is similar. Extract past event time series data of the international sports tournament shown in (a).

When similar past event time-series data is extracted by the similar event determination unit 132, the fluctuation demand prediction unit 133 uses the demand fluctuation amount and the event influence degree of the similar past event time-series data to perform real-time demand. Predict the amount of variation. This prediction is performed by the following formulas (3) and (4) using arbitrary constants γ and δ, for example.
(Real-time demand fluctuation) =
(Past demand fluctuation) x (Event influence rate change ratio) x γ + δ Equation (3)
(Event influence rate change rate) =
(Real-time event influence degree) / (Past event influence degree) Expression (4)
Note that γ and δ may be determined in advance based on statistical information such as the results of past predicted demand and actual demand, for example.

  Further, when predicting the real-time demand fluctuation amount, for example, it may be calculated on the assumption that the real-time demand fluctuation amount also changes in the same manner as the rate of change of the event influence degree with time in the past event time-series data. . Further, in consideration of the time zone of the event, for example, a matter that the demand fluctuation amount is large at noon and the demand fluctuation amount is small at night may be reflected.

  The fluctuation demand prediction unit 133 calculates a demand prediction result considering an event by adding a real-time demand fluctuation amount to the normal demand calculated by the normal demand prediction unit 131.

  Next, the flow of power demand prediction processing will be described. FIG. 14 is a flowchart illustrating a power demand prediction process performed by the power demand prediction apparatus 1 according to the first embodiment. Here, it is assumed that the daily demand consists of 48 frames of 30 minutes. As an example, it is assumed that the information source uses SNS Twitter (registered trademark).

  First, the event table generating unit 111 of the data organizing unit 11 inputs event data that is data related to an event (step S1), and at least a property related to the event including an event name, an event occurrence time, and an end time, An event table (FIG. 4) having “” is generated (step S2). In the event properties shown in FIG. 4, the “information sender” item includes the account name of the information sender on Twitter.

  Next, the event impact calculation unit 112 of the data organizing unit 11 calculates the event impact based on information acquired from any one or more of television, radio, and the Internet regarding the event (step S3). The event influence degree calculation unit 112 generates an event table with event influence degree (FIG. 5) by adding the event influence degree in association with the event to the event table. As described above, the event influence level is calculated from, for example, a poster's influence level and a poster's reliability level. In the case of Twitter, for example, the influence level of a poster can be calculated from the number of followers, the number of retweets of an article, and the like. Further, the reliability of the poster can be calculated from the credibility of information provided by the poster in the past, the difference between the general account and the official account, and the like.

  Next, the similarity determination unit 113 of the data organizing unit 11 determines the similarity between events based on one or more of the event name and property of the event table with event influence level (FIG. 5) ( Step S4), generating event time-series data (FIG. 7) having an event influence level for each predetermined unit time based on an event influence level associated with each of a plurality of similar events (step S4). S5).

  Next, the demand fluctuation determination unit 121 of the event determination unit 12 inputs the demand fluctuation data for each prediction target area, and determines the presence or absence of the demand fluctuation, that is, the time (time zone) when the demand fluctuation occurs and the demand fluctuation amount. (Step S6).

  Next, the event data extraction unit 122 of the event determination unit 12 determines the power based on the demand fluctuation amount that is the difference between the predicted demand and the actual demand calculated from the event time series data without considering the influence of the event. Event time-series data that affects demand is extracted (step S7), and the event time-series data and the demand fluctuation amount are associated with each other and stored as past event time-series data.

  Next, the normal demand prediction unit 131 of the demand prediction unit 13 predicts the normal demand for each prediction target area from the weather data (step S8).

  Next, the similar event determination unit 132 of the demand prediction unit 13 acquires or generates real-time event time-series data having an event influence degree for each predetermined unit time, and from a plurality of past event time-series data. Then, it is determined whether there is something similar to the real-time event time-series data (step S9).

  Next, the fluctuation demand prediction unit 133 of the demand prediction unit 13 predicts a real-time demand fluctuation amount using the demand fluctuation amount and event influence degree of similar past event time-series data (FIG. 13A). (Step S10). A plurality of past event time series may be used. In that case, what is necessary is just to cope by calculating the past demand fluctuation amount from an average value etc.

  Next, the fluctuation demand prediction unit 133 calculates a demand prediction result in consideration of the event by adding the real-time demand fluctuation amount calculated in Step S10 to the normal demand calculated in Step S8 (Step S11). ).

  Thus, according to the power demand prediction apparatus 1 of 1st Embodiment, the precision of the prediction of the future power demand which considered the event can be improved rather than before. For example, if there is an event where many people move to another place for several days, such as homecoming, there will be places where demand for ordinary household customers decreases and increases depending on the area. Also, if there are events where many people go out and watch TV, such as famous international sports competitions in foreign countries, the demand of ordinary household customers will increase. Considering the impact of such an event, future power demand can be predicted with high accuracy.

  Further, it is possible to accurately determine which past event affects the demand to the same extent as the influence on the demand by the event to be considered. In other words, it is possible to avoid using past event data that has the same event name but a different degree of influence on demand. Even if the event names are different, data of past events having the same (near) influence on demand can be used.

  Further, by using the event influence degree as described above, highly effective demand prediction can be performed.

(Second Embodiment)
Next, the power demand prediction apparatus 1 of 2nd Embodiment is demonstrated. In 2nd Embodiment, the event from which the magnitude | size of the influence regarding an electric power demand differs for every area as an event is assumed. An example of such an event is a fireworks display. In the case of a fireworks display, for example, the area closer to the fireworks display is considered to have a greater influence on demand. In addition, about the matter similar to 1st Embodiment, the overlapping description is abbreviate | omitted suitably and an item different from 1st Embodiment is demonstrated.

  FIG. 15 is a diagram illustrating an event table according to the second embodiment. FIG. 16 is a diagram illustrating a functional configuration of the event determination unit 12 of the power demand prediction apparatus 1 according to the second embodiment. As shown in FIG. 15, in the event table of the second embodiment, an event occurrence (holding) area and an information distribution area are added as properties of the event to the event table (FIG. 4) of the first embodiment. ing. Accordingly, an event distance calculation unit 123 is added to the event determination unit 12 as shown in FIG. The event distance calculation unit 123 calculates an event distance that is an index value of the magnitude of the influence on the power demand due to the event with respect to the prediction target area based on predetermined attribute information between the prediction target area and the event occurrence area (details). Will be described later).

  Hereinafter, the power demand prediction process by the power demand prediction apparatus 1 of 2nd Embodiment is demonstrated along the flowchart of FIG.

  First, the event table generation unit 111 inputs event data that is data related to an event (step S1), and generates an event table (FIG. 15) having an event name and a property (step S2).

  Next, the event influence degree calculation unit 112 calculates the event influence degree based on information acquired from any one or more of television, radio, and the Internet regarding the event (step S3). The event influence degree may be calculated based on the influence degree of the event holding place in addition to the influence degree of the poster and the reliability of the poster described in the first embodiment. The degree of influence of the event location includes the population and age ratio of the location where the event is held, the amount of traffic from other areas, and the like. Steps S4 and S5 are the same as in the first embodiment.

In step S6, the demand fluctuation determination unit 121 inputs demand fluctuation data for each prediction target area, and determines the presence or absence of demand fluctuation, that is, the time (time zone) when the demand fluctuation occurs and the demand fluctuation amount. Thereafter, the event distance calculation unit 123 calculates the event distance D between the event occurrence area and the surrounding area where the demand fluctuates due to the event. The event distance D _AB between area A and area B is defined by the following equation (5).

D _AB = f (Euclidean distance between areas, number of people moving between areas, area area (sum of two areas, product, etc.), population within area, age composition within area, number of contracted customers within area, area _AB (Including the number of train / bus commuter pass holders, the amount of demand fluctuation in the area due to past events, etc.) (5)

  That is, the event distance calculation unit 123 uses the function having at least one of the information in () on the right side of the equation (5) (predetermined attribute information between the prediction target area and the event occurrence area) as a variable. The distance D is calculated. In addition, what is necessary is just to use the information in () of the right side of Formula (5) in any or arbitrary combinations.

  In addition, for example, the Euclidean distance between areas is based on the distance from the event venue to the major cities in the area, the distance between the center and major cities of each area, the distance between locations such as prefectural offices, city halls, town halls, etc. calculate. The number of people moving between areas is calculated based on the number of vehicles traveling, the transport capacity of railways and buses, the number of main roads, and the like.

  Next, in step S <b> 7, the event data extraction unit 122 first determines an event held in the prediction target area for a time zone in which it is determined that there has been a demand fluctuation (which may include a time zone before and after that). If there is, the event time-series data is referred to, and if there is event time-series data that affects the power demand, it is extracted and stored as past event time-series data. This process is performed with reference to event time-series data of events having a short event distance D in order. Note that it is not necessary to process the event time-series data of events whose event distance D is longer than a certain level.

  In general, it is considered that the event occurrence area is most susceptible to demand fluctuation due to the event, so events that do not cause demand fluctuation in the event occurrence area are rejected as having no impact on demand fluctuation. May be. Further, the event distance D may be updated using the demand fluctuation in the event occurrence area and the demand fluctuation in the prediction target area. Step S8 is the same as that in the first embodiment.

  Next, in step S <b> 9, the similar event determination unit 132 determines whether there is something similar to the real-time event time-series data in order from the shortest event distance D from the event occurrence area among the past event time-series data. Judging. Step S10 is the same as that in the first embodiment.

  Next, in step S11, the fluctuation demand prediction unit 133 calculates the demand prediction result by adding the demand fluctuation amount calculated by the following equation (6) to the normal demand calculated in step S8 (step S11). S11).

(Demand fluctuation amount outside the event occurrence area) =
D _AB × (demand fluctuation amount in area A) × ε + ζ Equation (6)
Note that ε and ζ are arbitrary constants, and may be determined in advance based on, for example, predetermined past statistical information.

  In this way, according to the power demand prediction apparatus 1 of the second embodiment, it is possible to predict the future power demand with high accuracy in consideration of the event and the distance from the event occurrence area. That is, it is possible to predict the future power demand with high accuracy in response to an event that locally affects the event.

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

  For example, the time unit of future power demand prediction is not limited to a unit of 30 minutes, and may be another unit such as a unit of several hours or a unit of several days.

  Further, the data organizing unit 11 in the power demand prediction apparatus 1 may be realized by another PC (Personal Computer). In that case, the power demand prediction apparatus 1 acquires and uses event time-series data from the other PC.

1 Power demand forecasting device 2 CPU
3 Memory 4 Hard disk 5 GUI
6 Interface 7 Bus 11 Data Arrangement Unit 111 Event Table Generation Unit 112 Event Influence Level Calculation Unit 113 Similarity Determination Unit 12 Event Determination Unit 121 Demand Fluctuation Determination Unit 122 Event Data Extraction Unit 123 Event Distance Calculation Unit 13 Demand Prediction Unit 131 Normal Demand Prediction unit 132 Similar event determination unit 133 Fluctuation demand prediction unit

Claims (8)

It is indicated by the predicted demand calculated from the event time series data that has the event influence degree indicating the influence degree of the electric power demand every predetermined unit time without considering the influence of the event and the predicted demand. Based on the demand fluctuation amount that is the difference from the actual demand that is the actual power demand that occurred in the specified time zone, the event time series data that affected the power demand is extracted, and the event time series data and the demand fluctuation An event determination unit that associates the amount with each other and stores it as past event time-series data;
A normal demand prediction unit that predicts normal demand for each area to be predicted without considering the influence of the event;
Obtain or generate real-time event time-series data having an event influence degree indicating the influence degree of a real-time event on the power demand for each predetermined unit time, and from the plurality of the past event time-series data, A similar event determination unit that determines whether there is something similar to real-time event time-series data;
Demand considering the event by predicting the real-time demand fluctuation amount using the demand fluctuation amount and event influence degree of the similar past event time-series data, and adding the normal demand to the demand fluctuation amount A power demand prediction device comprising: a fluctuation demand prediction unit that calculates a prediction result. Event data that is data related to the event, and an event table generation unit that generates an event table having at least an event name, an event occurrence time, and a property related to the event including an end time;
With respect to the event, the event influence degree is calculated based on information acquired from at least one of television, radio, and the Internet, and the event influence degree is added to the event table in association with the event. An event impact calculator that generates an event table with an impact,
Based on any one or more of event names and properties in the event table with event influence degree, the similarity degree between the events is determined, and the event influence degree associated with each of a plurality of similar events is determined. The power demand prediction apparatus according to claim 1, further comprising: a similarity determination unit that generates the event time-series data having the event influence degree for each predetermined unit time.   The power demand prediction apparatus according to claim 1, wherein the event influence degree is an index indicating the number of people who are affected by an action when the event occurs. The event determination unit
From the event time series data, for the time zone in which it is determined that there has been a demand fluctuation based on the demand fluctuation amount, the event time series data of the event that has affected the power demand is extracted, and the event time series data and The power demand prediction apparatus according to claim 1, further comprising an event data extraction unit that associates the demand fluctuation amount with each other and stores the data as past event time-series data.   The event data extraction unit includes a similarity between a time zone in which a demand fluctuation is determined and an occurrence time zone of the event, and a waveform of a demand fluctuation and a fluctuation waveform of an event influence degree of the event time series data. The power demand prediction apparatus according to claim 4, wherein the event time-series data of an event that has influenced power demand is extracted based on at least one of the similarities.   The similar event determination unit determines whether there is a similarity to the real-time event time-series data from a plurality of the past event time-series data, a time zone in which it is determined that there is a demand fluctuation, and an occurrence time of the event The power demand prediction apparatus according to claim 1, wherein determination is made based on at least one of a band similarity and a similarity between a waveform of demand fluctuation and a waveform of event influence degree fluctuation of the event time-series data. . The event is an event having a different magnitude of influence on power demand for each area,
The event determination unit calculates an event distance that is an index value of the magnitude of the influence on the power demand by the event with respect to the prediction target area based on predetermined attribute information between the prediction target area and the event occurrence area A distance calculation unit;
The power demand prediction apparatus according to claim 1, wherein the fluctuation demand prediction unit predicts a real-time demand fluctuation amount according to the event distance.   The event distance calculation unit includes the event distance, the Euclidean distance between areas, the number of people moving between areas, the area area, the population in the area, the age composition in the area, the number of contracted customers in the area, and the railway / bus including both areas. The power demand prediction apparatus according to claim 7, wherein the power demand forecasting device calculates the number of commuter pass holders and a function having at least one of the in-area demand fluctuation amount due to a past event as a variable.

Images