JP2017220980A - Prediction apparatus, prediction method and prediction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a prediction apparatus, a prediction method and a prediction program capable of accurately predicting power demand to an electric power company.SOLUTION: A generation unit 41 predicts power demand at predetermined prediction target time by using power demand results data 30 obtained up to prediction execution time. A prediction unit 42 predicts power demand at prediction target time from the power demand at the prediction execution time by using a generated prediction model. An update instruction unit 44 instructs update of the prediction model to the generation unit 41 at predetermined timing.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a prediction apparatus, a prediction method, and a prediction program.

  With the increase in renewable energy equipment such as photovoltaic power generation equipment installed on the customer side, it is difficult for power companies to predict power demand. The power company can accurately grasp the power demand for the power company (hereinafter also referred to as “net demand”), which is the true power demand minus the generated power of the renewable energy equipment installed on the customer side. It is only a value. Power companies can't accurately measure both true power demand and the power generated by renewable energy equipment. For this reason, with the increase in the renewable energy equipment installed on the consumer side, it is increasingly difficult to accurately predict Internet demand.

  Therefore, a technique for predicting power demand by regression analysis has been proposed. For example, the electric power demand is predicted by regression analysis using a forecast formula using weather forecast values such as temperature and solar radiation amount and actual values of electric power demand as explanatory variables.

JP 2014-54048 A JP 2011-2929 A

  In a conventional demand prediction method based on regression analysis, weather forecast values such as temperature and solar radiation are used as explanatory variables used in the prediction formula. However, it is not always easy to obtain highly accurate weather forecast values. For this reason, the electric power demand with respect to an electric power company may be unable to be estimated accurately.

  This invention is made | formed in view of the above, Comprising: It aims at providing the prediction apparatus, the prediction method, and prediction program which can estimate the electric power demand with respect to an electric power company accurately.

  In order to solve the above-described problems and achieve the object, the prediction device of the present invention predicts the power demand at a predetermined prediction target time using the actual power demand data obtained up to the time when the prediction is performed. Using the prediction model generated by the generation unit, a prediction unit for predicting the power demand at the prediction target time from the power demand at the time of prediction execution, and predicting the generation unit at a predetermined timing And an update instruction unit for instructing to update the model.

  The present invention has an effect of accurately predicting power demand for an electric power company.

FIG. 1 is a diagram illustrating an example of the current day correction prediction. FIG. 2 is a diagram showing air conditioning effect variables. FIG. 3 is a diagram illustrating an example of a functional configuration of the prediction apparatus. FIG. 4 is a diagram illustrating an example of a time-series change in the PV connection amount. FIG. 5 is a diagram illustrating an example of the verification result of the prediction accuracy by the prediction method according to the first embodiment. FIG. 6 is a diagram showing the number of occurrence days for each month when the prediction is significantly different at any time. FIG. 7 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the prediction is significantly out of the power failure due to typhoon damage. FIG. 8 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the power failure occurs. FIG. 9 is a diagram illustrating a change in the actual value and the predicted value of the power demand on the day when the sudden change in the amount of sunshine occurs. FIG. 10 is a diagram showing changes in the amount of solar radiation. FIG. 11 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when a sudden change in temperature occurs. FIG. 12 is a diagram showing changes in temperature. FIG. 13 is a diagram showing changes in the actual value and predicted value of the power demand on the day when a sudden change in temperature and solar radiation occurs. FIG. 14 is a diagram showing changes in temperature. FIG. 15 is a diagram showing changes in the amount of solar radiation. FIG. 16 is a flowchart illustrating an example of the procedure of the prediction process. FIG. 17 is a diagram illustrating an example of a verification result of prediction accuracy by the prediction method according to the second embodiment. FIG. 18 is a diagram showing the number of days of occurrence for each month when the prediction is significantly different at any time. FIG. 19 is a diagram illustrating the change in the actual value and the predicted value of the power demand on the day when the prediction deviates greatly due to a power failure due to typhoon damage. FIG. 20 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the sudden change in the amount of sunshine occurs. FIG. 21 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when a sudden change in temperature occurs. FIG. 22 is a diagram showing changes in the actual value and the predicted value of the power demand on the day when the temperature and the amount of solar radiation suddenly change. FIG. 23 is a flowchart illustrating an exemplary procedure of a prediction process according to the second embodiment. FIG. 24 is a diagram illustrating a computer that executes a prediction program.

  Hereinafter, embodiments of a prediction apparatus, a prediction method, and a prediction program according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments. Each embodiment can be appropriately combined within a range in which processing contents are not contradictory.

[Prediction method]
First, the prediction method according to the first embodiment will be described. Below, the case where the electric power company estimates the net demand of the target area which supplies electric power is demonstrated to an example. An electric power company supplies electric power generated by a power generator such as a thermal power plant to each consumer in the target area through a distribution system.

  By the way, as mentioned above, the electric power company cannot measure an accurate value of the true power demand and the generated power of the renewable energy equipment. Therefore, it is conceivable to predict net demand by regression analysis using prediction formulas that use weather forecast values such as temperature and solar radiation and actual values of power demand as explanatory variables. However, it is not always easy to obtain highly accurate weather forecast values. For example, it is not always easy to obtain highly accurate weather forecast values for the amount of solar radiation that is important for predicting net demand. For this reason, the electric power demand with respect to an electric power company may be unable to be estimated accurately.

  Therefore, in the prediction method according to the first embodiment, the net demand is predicted using only the actual value obtained up to the prediction execution time point. A major feature of the prediction method according to the first embodiment is that prediction is performed with high accuracy by using various actual values available at the time of prediction without using uncertain future forecast values. Further, in the prediction method according to the first embodiment, a prediction model used for prediction of net demand is updated in a short period.

  By the way, various studies for the purpose of accurately predicting Internet demand have been conducted. Such research is roughly divided into the following two methods. One method is to predict the generated power of the renewable energy equipment installed on the consumer side and the true power demand, respectively, and subtract the generated power of the renewable energy equipment from the predicted true power demand to Is a method for predicting The other method is a method for directly predicting net demand without particularly predicting true power demand. The prediction method according to the first embodiment is a method for directly predicting net demand. Hereinafter, the net demand forecast is referred to as demand forecast.

  First, a case where the net demand is predicted with high accuracy using the prediction method according to the first embodiment will be described with reference to a prediction work called same-day correction prediction performed at an electric power company. In the current day correction forecast, the demand from 30 minutes ahead to 4 hours 30 minutes ahead is predicted based on the latest performance data at 4 time points (5:30, 8:30, 10:30, 14:30) in the daytime. To do. The reason for considering the revised forecast on the day is due to the following two reasons. One reason is that the number of renewable energy equipment installed on the consumer side is expected to increase in the future, and as a result, it becomes difficult to accurately forecast demand ahead for a long time, and the importance of the current day forecast will increase. This is because it is considered. Another reason is that, since April 2016, the establishment of a market for electricity one hour ago is being considered, and it is thought that demand forecasts in a short time will become increasingly important.

  The electric power company is reviewing the forecast in the time zone from 6:00 to 18:00 when the error of the previous day's forecast becomes large due to the influence of the amount of solar radiation, etc., and this is called the current day forecast. In the current day correction prediction, for example, power demand ahead of about 30 minutes to 5 hours is predicted four times a day. FIG. 1 is a diagram illustrating an example of the current day correction prediction. In the example of FIG. 1, the electric power demand of 6: 0 to 8:30 is predicted at the time of 5:30 as the first time. As the second time, the power demand between 9:00 and 10:30 is predicted at 8:30. Moreover, the electric power demand of 11: 00 to 14:30 is predicted as the third time at 10:30. In addition, as the fourth time, the power demand from 15:00 to 19:00 is predicted at 14:30.

  A prediction model used for the prediction method according to the first embodiment will be described. In the prediction method according to the first embodiment, a prediction model that predicts the power demand at a predetermined prediction target time is generated using the actual power demand data obtained up to the time when the prediction is performed. In the prediction method according to the first embodiment, the power demand at the prediction target time is predicted from the power demand at the time of the prediction using the generated prediction model. For example, in the current day corrected prediction performed at 5:30, the net demand at time t is predicted using the following prediction formula (1) as a prediction model.

Net demand at time t = a ₁ × average demand at time t
+ A ₂ × Deviation in demand at the time of forecasting
+ A ₃ × Saturday dummy variable
+ A ₄ × Sunday / holiday dummy variable
+ A ₅ × air conditioning effect variable
+ A ₆ × 1 year ago dummy variable
+ A ₇ × 2 years ago dummy variable
+ A ₈ (1)

  In addition, for example, in the current day correction prediction performed at 8:30, 10:30, and 14:30, the net demand at time t is predicted using the following prediction formula (2) as a prediction model.

Net demand at time t = a ₁ × average demand at time t
+ A ₂ × Deviation in demand at the time of forecasting
+ A ₃ × Saturday dummy variable
+ A ₄ × Sunday / holiday dummy variable
+ A ₅ × air conditioning effect variable
+ A ₆ x solar radiation influence variable
+ A ₇ × 1 year ago dummy variable
+ A ₈ × 2 years ago dummy variable
+ A ₉ (2)

  Here, the reason why the prediction formula (1) is used only for the current day correction prediction carried out at 5:30 is that the amount of solar radiation is 0 at the prediction time of 5:30 before sunrise.

Details of the explanatory variables used in the prediction equations (1) and (2) will be described below.
Average demand at time t: This is the average power demand at time t in a predetermined period (for example, 49 days) that is closest to the prediction execution time.
Deviation in demand at the time of forecasting: The difference between the average demand at the time of forecasting and the actual value of demand.
Saturday dummy variable: A variable that takes a value of 1 if the prediction target day is a Saturday other than a holiday, and 0 otherwise.
Air conditioning effect variable: A value calculated by the following equation (3).

  Heating / cooling effect variable = max (d-20; 18-d; 0) (3)

  Here, d is the actual value of the temperature at the time of prediction execution. max (d-20; 18-d; 0) is a function that outputs the maximum value among d-20, 18-d, and 0. FIG. 2 is a diagram showing air conditioning effect variables. The air conditioning effect variable increases in proportion to the temperature when the temperature d exceeds 20 ° C., and increases in inverse proportion to the temperature when the temperature d falls below 18 ° C. The air conditioning effect variable increases in proportion to the temperature when the temperature d exceeds 20 ° C., thereby reflecting the change in cooling demand. When the temperature d falls below 18 ° C., the value increases in inverse proportion to the temperature. This reflects changes in heating demand.

  Solar radiation influence variable: It is a value calculated by the following formula (4).

  Solar radiation influence variable = w x solar radiation amount at the time of forecast implementation (4)

  Here, w is a coefficient for correcting the difference in the influence of solar radiation on the power demand due to the difference in the connection amount of the photovoltaic power generation facilities in each month. For the coefficient w, for example, a larger value is set for a month with a larger amount of connection of the solar power generation equipment in correspondence with the connection amount of the solar power generation equipment of each month. For example, since the connection amount of photovoltaic power generation facilities has increased in recent years, the coefficient w is set to a larger value as the month when the prediction execution time is newer.

One-year-old dummy variable: A variable that takes a value of 1 if the past data used for the coefficient of the prediction formula is one year before the year to be predicted, and 0 otherwise.
Dummy variable 2 years ago: A variable that takes a value of 1 if the past data used for the coefficient of the prediction formula is 2 years before the year to be predicted, and 0 otherwise.

In the prediction method according to the first embodiment, a prediction model that predicts the power demand at a predetermined prediction target time is generated using the actual power demand data obtained up to the time when the prediction is performed. For example, in the current day correction prediction carried out at 5:30, the values of the coefficients (a _{1 to} a ₈ ) of the prediction formula (1) are determined using the performance data for the past two years. In the day-corrected predictions at 8:30, 10:30, and 14:30, the values of the coefficients (a _{1 to} a ₉ ) of the prediction formula (2) are determined using the actual data for the past two years. To do. Specifically, the past four weeks of the year from the prediction execution time, four weeks before and after the prediction target date one year ago (for example, 336 days to 392 days before the prediction execution date) The coefficient value of the prediction formula is determined so that the prediction error is minimized with respect to the performance data for a week (for example, 700 days to 756 days before the prediction execution date).

Thereby, the coefficient a ₁ of the prediction equations (1) and (2) becomes a value corresponding to the component of the power demand related to the average demand at time t. The coefficient a ₂ of the prediction equations (1) and (2) is a value corresponding to the component of the power demand related to the demand deviation at the time of the prediction execution. The coefficient a ₃ of the prediction formulas (1) and (2) is a value corresponding to a component of power demand peculiar to Saturday. The coefficient a ₄ of the prediction equations (1) and (2) is a value corresponding to the component of power demand peculiar to Sunday. The coefficient a ₅ in the prediction equations (1) and (2) is a value corresponding to the component of the power demand for air conditioning. The coefficient a ₆ of the prediction formula (2) is a value corresponding to a decrease in power demand according to the power generated by renewable energy equipment such as solar power generation equipment. Coefficient a ₇ of the coefficient a ₆ and a prediction equation of the prediction equation (1) (2) is a value corresponding to the component of the specific power demand a year ago. Coefficients a ₈ coefficients a ₇ and a prediction equation of the prediction equation (1) (2) is a value corresponding to the component of the specific power demand two years ago. Coefficients a ₉ coefficients a ₈ and a prediction equation of the prediction equation (1) (2) is a value corresponding to the base of the components included in the total power demand.

[Configuration of prediction device]
Next, the configuration of the prediction device 10 to which the above prediction method is applied will be described. The prediction device 10 is a device that performs prediction using the above-described prediction method. In addition, a present Example demonstrates the case where electric power demand is predicted by the prediction apparatus 10 using the above-mentioned prediction method. The prediction apparatus 10 is an information processing apparatus such as a server computer, a desktop PC (personal computer), a tablet PC, or a notebook PC. The prediction device 10 may be implemented as a single computer, or may be implemented as a cloud of a plurality of computers. In the present embodiment, a case where the prediction apparatus 10 is a single computer will be described as an example.

  FIG. 3 is a diagram illustrating an example of a functional configuration of the prediction apparatus. As illustrated in FIG. 3, the prediction device 10 includes a display unit 20, an input unit 21, a storage unit 22, and a control unit 23. The prediction device 10 may have various known functional units in addition to the functional units illustrated in FIG. For example, the prediction device 10 may include a communication interface unit that performs communication with other terminals.

  The display unit 20 is a display device that displays various types of information. Examples of the display unit 20 include a display device such as an LCD (Liquid Crystal Display). The display unit 20 displays various information. For example, the display unit 20 displays various operation screens and prediction results.

  The input unit 21 is an input device that inputs various types of information. For example, examples of the input unit 21 include an input device such as a keyboard and a mouse connected to the prediction device 10, various buttons provided on the prediction device 10, and a transmissive touch sensor provided on the display unit 20. . In the example of FIG. 3, since the functional configuration is shown, the display unit 20 and the input unit 21 are separately provided. For example, the display unit 20 and the input unit 21 such as a touch panel are integrally configured. May be.

  The storage unit 22 is a storage device that stores various data. For example, the storage unit 22 is a storage device such as a hard disk, an SSD (Solid State Drive), or an optical disk. The storage unit 22 may be a semiconductor memory that can rewrite data such as a random access memory (RAM), a flash memory, and a non-volatile static random access memory (NVSRAM).

  The storage unit 22 stores an OS (Operating System) executed by the control unit 23 and various programs. For example, the storage unit 22 stores various programs including a prediction program that executes a prediction process described later. Furthermore, the storage unit 22 stores various data used in programs executed by the control unit 23. For example, the storage unit 22 stores performance data 30.

  The performance data 30 is data storing performance values related to various parameters used for prediction of power demand. For example, in the performance data 30, the amount of power supplied to the target area to which power is supplied by the power company that is the target of net demand is stored as the net demand value. For example, the performance data 30 stores the actual values for each date and time for a predetermined period in the past of the net demand, temperature, and amount of sunshine for the target area. The predetermined period may be longer than the period of data used for generating the prediction model. The temperature and the amount of sunlight may be measured values measured at any one point in the target area, for example, and may be an average value or a weighted average value of measured values measured at a plurality of points in the target area. There may be. For example, when using measured values measured at multiple points in the target area for the temperature and sunshine amount, the target area is divided according to each point, and the various areas are divided according to the power generation rate of the divided divided areas. You may use the average value which weighted the measured value of the point.

  The control unit 23 is a device that controls the prediction device 10. As the control unit 23, an electronic circuit such as a CPU (Central Processing Unit) and an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) and an FPGA (Field Programmable Gate Array) can be employed. The control unit 23 has an internal memory for storing programs defining various processing procedures and control data, and executes various processes using these. The control unit 23 functions as various processing units by operating various programs. For example, the control unit 23 includes an acquisition unit 40, a generation unit 41, a prediction unit 42, an output control unit 43, and an update instruction unit 44.

  The acquisition unit 40 acquires various data. For example, the acquisition unit 40 acquires actual values of various parameters used for power demand prediction. For example, the acquisition unit 40 acquires the net demand at each date and time from the information provider device that stores the net demand actual value for each time of the power company to be predicted for the net demand. In addition, the acquisition unit 40 acquires the temperature and sunshine amount of each date and time from the information source device that stores the temperature and sunshine amount for each time of the target area to which the electric power company that is the target of network demand supplies power. To do. The network demand, temperature, and amount of sunshine are measured based on the latest measured date and time data of the network demand, temperature, and amount of sunlight in response to a request from the acquisition unit 40. You may send with. Further, as for the net demand, the temperature, and the amount of sunshine, the information source device may periodically transmit the data of the net demand, the temperature, and the amount of sunshine of the latest measured date and time together with the measured date and time.

  The acquisition unit 40 stores the actual values of various parameters used for the prediction of the acquired power demand in the actual data 30. For example, the acquisition unit 40 stores the net demand, the temperature, and the amount of sunlight in the performance data 30 in association with the measured date and time.

The production | generation part 41 produces | generates the prediction model which estimates the electric power demand of the time of predetermined | prescribed prediction object using the performance data 30 of the electric power demand obtained by prediction implementation time, temperature, and the amount of sunlight. For example, the generation unit 41 is instructed by the update instruction unit 44 to update the prediction model at 5:30, 8:30, 10:30, and 14:30 that are prediction execution times of the current day correction prediction. The generation unit 41 determines the values of the coefficients (a _{1 to} a ₈ ) of the above-described prediction formula (1) using the actual data 30 for the past two years in the correction correction prediction of 5:30 on the day. For example, the generation unit 41 predicts power demand at 6:00, 6:30, 7:00, 7:30, 8:00, and 8:30 every 30 minutes from 5:30 to 8:30. In this case, 6:00, 6:30, 7:00, 7:30, 8:00, and 8:30 are set as the prediction target time t, and the prediction formula (1) is generated for each prediction target time t. For example, the generation unit 41 sets a value based on the performance data 30 to each parameter of the prediction formula (1) described above from the performance data 30 for each prediction target time t, and the prediction error of the prediction formula (1) is larger. The values of the coefficients (a _{1 to} a ₈ ) are determined so as to be the smallest. For example, the generation unit 41 uses the performance data 30 for each of the four weeks before and after the prediction target date of one year ago for the past four weeks and the four weeks before and after the prediction target date of two years ago from the prediction execution time. Prediction formula (1) when setting average demand at time t, deviation of demand at the time of forecasting, Saturday dummy variable, heating / cooling effect variable, dummy variable for one year, dummy variable for two years ago for each day The regression analysis is performed so as to reduce the prediction error between the predicted value of the demand at time t obtained from the actual value of the demand at time t, and the values of the coefficients (a _{1 to} a ₈ ) are determined.

In addition, the generation unit 41 uses the actual data 30 for the past two years in the 8:30, 10:30, and 14:30 on-site correction predictions, and uses the coefficients (a _{1 to} a ₉ ) of the prediction formula (2). Determine the value of. For example, the generation unit 41 uses the actual data 30 for the past two years in the current day correction prediction performed at 8:30, and uses the values of the coefficients (a _{1 to} a ₉ ) of the prediction formula (2) described above. To decide. For example, when the generation unit 41 predicts the power demand of 9:30, 10:00, 10:30 every 30 minutes from 9:30 to 10:30, 9:30, 10:00, 10: The prediction formula (2) is generated for each prediction target time t, with 30 as the prediction target time t. For example, the generation unit 41 sets a value based on the performance data 30 to each parameter of the prediction formula (2) described above from the performance data 30 for each prediction target time t, and the prediction error of the prediction formula (2) is larger. The values of the coefficients (a _{1 to} a ₉ ) are determined so as to be the smallest. For example, the generation unit 41 uses the performance data 30 for each of the four weeks before and after the prediction target date of one year ago for the past four weeks and the four weeks before and after the prediction target date of two years ago from the prediction execution time. For each day, the average demand at time t, the deviation in demand at the time of forecasting, the Saturday dummy variable, the heating / cooling effect variable, the solar radiation effect variable, the one-year-old dummy variable, and the forecast when setting the two-year-old dummy variable Regression analysis is performed so that the prediction error between the predicted value of demand at time t obtained from Equation (2) and the actual value of demand at time t is reduced, and the values of the coefficients (a _{1 to} a ₉ ) are determined.

  That is, the generation unit 41 uses the power demand, temperature, and sunshine amount actual data 30 obtained up to the prediction execution time as a prediction model at each prediction target time t at the prediction execution time. ) Or the prediction formula (2) is generated.

  The prediction unit 42 uses the prediction model generated by the generation unit 41 to predict the power demand at the time to be predicted from the power demand, temperature, and amount of sunlight at the time of the prediction. For example, the prediction unit 42 obtains the average demand at the time t at the prediction execution time and the deviation of the demand at the prediction execution time from the result data 30. Moreover, the prediction part 42 calculates | requires an air conditioning effect variable using Formula (3) from the temperature at the time of prediction implementation. Moreover, the prediction part 42 calculates | requires a solar radiation influence variable using Formula (4) from the solar radiation amount at the time of prediction implementation. The prediction unit 42 sets a Saturday dummy variable and a Sunday / holiday dummy variable from the date and day of the prediction. The prediction unit 42 sets 0 for both the one-year-old dummy variable and the two-year-old dummy variable. In addition, when the prediction implementation time is 5:30, the solar radiation influence variable may not be obtained.

  The prediction unit 42 sets each parameter in the prediction formula for each prediction target time t, and calculates the net demand at the time t.

  The output control unit 43 controls various outputs. For example, the output control unit 43 causes the display unit 20 to display the information on the net demand for each prediction target time t calculated by the prediction unit 42 as a prediction result. Further, for example, the output control unit 43 outputs information on net demand for each prediction target time t to an external terminal device. Thereby, the electric power company can grasp | ascertain the prediction result of the net demand for every time t of prediction object from the output information.

  The update instruction unit 44 instructs the generation unit 41 to update the prediction model at a predetermined timing. For example, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at a timing when the electric power company reviews the operation plan of the generator. For example, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at 5:30, 8:30, 10:30, and 14:30 where the current day correction prediction is performed. In addition, although the update instruction | indication part 44 demonstrated as an example the case where an electric power company instruct | indicates the update of a prediction model with respect to the production | generation part 41 at the timing which reviews the operation plan of a generator, it updates the prediction model at other timings. May be indicated. For example, the update instruction unit 44 may instruct the generation unit 41 to update the prediction model at a predetermined time when the power demand tendency changes during the day. For example, the update instruction unit 44 may instruct the generation unit 41 to update the prediction model at times 6:00, 9:00, 12:00, and 17:00, respectively. Here, for example, 6:00 changes from night to morning, and since many people get up and start activities, the tendency of power demand changes. 9:00 is the time when a factory or the like starts, and the tendency of power demand changes. 12:00 is the time when the sun rises and the difference in the amount of sunshine due to the weather increases. The amount of power generated by the solar power generation device changes depending on the amount of sunshine, so the power demand trend changes. Moreover, 17:00 is the time when a factory or the like ends, and the tendency of power demand changes. In addition, although 6:00, 9:00, 12:00, and 17:00 were illustrated as predetermined time when the tendency of an electric power demand changes, it is not limited to this. In addition, for example, the update instruction unit 44, as will be described later, is a timing that is affected by a typhoon, a timing that a power failure occurs, a timing that a fluctuation of a predetermined amount or more occurs in the amount of solar radiation, and a timing that a fluctuation that exceeds a predetermined amount occurs in the temperature. The generation unit 41 may be instructed to update the prediction model at any timing. For example, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at a timing when the difference between the predicted value of demand predicted by the prediction unit 42 and the actual value of demand exceeds a certain value. May be. For example, the update instruction unit 44 is the time when the difference between the predicted value of the demand predicted by the prediction unit 42 and the actual value of the demand exceeds a certain value (for example, 2%) and the difference exceeds the certain value. For subsequent predicted times, an update of the prediction model may be instructed, and prediction of power demand at the predicted time may be executed again.

[Prediction example]
Next, an example of prediction by the prediction device 10 to which the above-described prediction method is applied will be described. Hereinafter, a case where the prediction accuracy is verified using the power demand result data 30 measured at intervals of 30 minutes from April 1, 2009 to February 28, 2015 will be described as an example. Since the prediction formulas (1) and (2) require the past two years of actual data 30 to determine the coefficient, here, from April 1, 2012 when the past two years of actual data 30 are available. On-the-day correction prediction until February 28, 2015 is performed, and the prediction accuracy is verified.

  Moreover, the prediction of power demand by the prediction formulas (1) and (2) requires actual values of the temperature at the time of prediction and the amount of solar radiation as data relating to the weather. Here, prediction is performed using the hourly temperature measured at the Meteorological Agency observation point closest to the head office of the electric power company that is the target of net demand, and the actual value of global solar radiation, and the prediction accuracy is verified. .

  The solar radiation amount correction value is determined as follows. Based on the actual value of the connection amount of the photovoltaic power generation (PV) installed on the customer side in each fiscal year, the connection amount of each month is complemented with a quadratic function approximation. FIG. 4 is a diagram illustrating an example of a time-series change in the PV connection amount. FIG. 4 shows the connection amount of each month normalized with the connection amount of March 2010 as 1.0. As shown in FIG. 4, in the target electric power company, the PV connection amount has increased rapidly since around 2012.

FIG. 5 is a diagram illustrating an example of the verification result of the prediction accuracy by the prediction method according to the first embodiment. The prediction error for each time in each year is as shown in FIG. 5, and the following tendency can be seen.
-In the morning time zone (6: 00 to 8:30) where prediction is easy with little fluctuation, the prediction error is large.
・ The forecast error is larger in the latest fiscal year.

  In the morning time zone, the level of demand is not high and the amount of solar radiation is not so large. Therefore, the fluctuation in daily demand is small compared to the daytime hours thereafter, and it is considered relatively easy to predict. However, according to the verification result, the prediction error in the morning time zone is the same as that in the daytime peak time zone. The reason for the large prediction error in the morning time zone may be that the predicted value in the morning time zone is a value that does not take into account the effects of solar radiation. In the prediction of 5:30 performed before sunrise, since the actual solar radiation amount value at the time of the prediction is 0, the prediction formula (1) that does not use the actual solar radiation amount value as an explanatory variable is performed. For this reason, the prediction of 6:00 to 8:30 is a prediction that does not reflect the difference in the amount of solar radiation on the day, and this is considered to lead to an increase in prediction error.

  Moreover, it is thought that the spread of PV is having a big influence as a cause that the prediction error becomes so large that it is the most recent degree in 2012, 2013, and 2014. As shown in FIG. 4, the PV connection amount is increasing rapidly every year, which is considered to affect the increase in prediction error.

Next, the frequency of occurrence and the cause of the prediction will be described. In the present embodiment, the case where the prediction error exceeds 10% is regarded as a great deviation. FIG. 6 is a diagram showing the number of occurrence days for each month when the prediction is significantly different at any time. Similar to the prediction accuracy verification results shown in FIG. 5, the number of days of occurrence of the predicted outlier in 2012 was the smallest, and the number of days in 2014 was the largest. In addition, when the cause of the deviation is examined, the following three are the main causes.
・ Effect of typhoon (power failure and reduction of power demand)
・ Power failure ・ Rapid change of weather

  In the following, detailed causes will be described by taking typical days as an example for each cause of a large deviation.

  First, a day on which a large deviation is considered to have occurred due to the typhoon will be described. The prediction error increases when power facilities are damaged by a typhoon and a power failure occurs. The forecast error also increases when the typhoon affects business activities, such as the company or factory being closed, and the demand pattern is significantly different from normal. The major outages in September 2012, October 2013, July 2014 and October shown in FIG. 6 are considered to be caused by the influence of the typhoon.

  Among these, the actual value and the predicted value of the power demand in September 2012 will be described. FIG. 7 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the prediction is significantly out of the power failure due to typhoon damage. In FIG. 7, the power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. Normally, the power demand continues to increase from 6:00 to 12:00. However, on this day, power demand continues to decrease from 6:00 due to the influence of the typhoon. On the other hand, in the prediction at 5:30, it is predicted that the demand will increase as usual after 6:00 based on the past results, so that the prediction is greatly out of place. In the second prediction conducted at 8:30 and the third prediction conducted at 10:30, the forecast value has been revised downward according to the actual value, but the power demand is significantly below the normal level. . For this reason, the prediction error remains unchanged. As a result of the fourth-day correction prediction performed at 14:30, the prediction error is finally reduced.

  Next, the day when it is considered that a major outage has occurred due to a power failure will be described. Of the days when the forecast deviates, one case in March 2013 has a large prediction error due to a short-term power outage caused by a failure of the power generation facility. A description will be given using actual values and predicted values of power demand on the day when a power failure occurs. FIG. 8 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the power failure occurs. In FIG. 8, the power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 8, on this day, a failure occurred in the power generation facility around 16:00, and a power failure occurred along with this, resulting in a large prediction error. However, the power failure is restored in a short time, and at the same time, the prediction error is restored to the normal level.

  Next, a description will be given of the day on which it is considered that a large deviation has occurred due to a sudden change in weather. In a case where the prediction was greatly deviated in December 2013, the prediction error was large due to a sudden change in the amount of sunlight as a sudden change in the weather. FIG. 9 is a diagram illustrating a change in the actual value and the predicted value of the power demand on the day when the sudden change in the amount of sunshine occurs. In FIG. 9, the power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 9, on this day, the demand has increased from the prediction at around 12:00, and the prediction error has increased. FIG. 10 is a diagram showing changes in the amount of solar radiation. FIG. 10 shows the change in the amount of solar radiation on the same day as FIG. In FIG. 10, the total solar radiation amount is normalized with a predetermined reference value of 1.0. As shown in FIG. 10, when the change in the total solar radiation amount is seen on this day, the solar radiation amount is rapidly decreasing around 12:00. Due to this influence, the generated power of PV decreased rapidly, and the rapid increase in net demand is considered to have caused the prediction to be out of place.

  Further, in another case in which the prediction deviates greatly in December 2013, the prediction error is large due to a sudden change in temperature as a sudden change in weather. FIG. 11 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when a sudden change in temperature occurs. In FIG. 11, the electric power demand is normalized with a predetermined reference value of 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 11, on this day, the prediction error is large until the fourth day of the current day correction prediction is performed at 14:00 and after 12:00. FIG. 12 is a diagram showing changes in temperature. FIG. 12 shows the change in temperature on the same day as FIG. FIG. 12 shows the difference between the average daily temperature of the target day and the temperature at each time. As shown in FIG. 12, on this day, the temperature suddenly rises in one hour from 10:00 to 11:00. The sudden rise in temperature caused a drastic decrease in heating demand and the rapid decrease in net demand.

  In addition, although the example where it is thought that the prediction was greatly deviated due to the sudden change of only the solar radiation amount or only the temperature was shown, there is a case where the sudden change of both the solar radiation amount and the temperature is considered to be the cause of the prediction deviation. FIG. 13 is a diagram showing changes in the actual value and predicted value of the power demand on the day when a sudden change in temperature and solar radiation occurs. In FIG. 13, the power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. The example of FIG. 13 shows a case that deviated greatly in February 2015. As shown in FIG. 13, on this day, the prediction error suddenly increases after 11:00. FIG. 14 is a diagram showing changes in temperature. FIG. 14 shows the temperature change on the same day as FIG. FIG. 14 shows the difference between the average daily temperature of the target day and the temperature at each time. As shown in FIG. 14, on this day, the temperature has risen sharply from 9:00 to 10:00. Moreover, FIG. 15 is a figure which shows the change of the total solar radiation amount. FIG. 15 shows the change in the amount of solar radiation on the same day as FIG. As shown in FIG. 15, the amount of global solar radiation increases rapidly in one hour from 9:00 to 10:00, similarly to the temperature. This day was affected by sudden changes in both temperature and solar radiation, and it is thought that the forecast was out of place.

  As for the sudden change in the weather, the number of times the temperature changes suddenly is smaller than that of the sudden change in the amount of solar radiation, and most of the causes of the deviation are due to the sudden change in the amount of solar radiation. In addition, the PV installed on the consumer side has increased, and the impact of changes in solar radiation on the net demand for electricity has been increasing in recent years. It is thought that.

  Therefore, for example, the update instructing unit 44 is one of the timing of being affected by a typhoon, the timing of a power failure, the timing of a fluctuation of a predetermined amount or more in the amount of solar radiation, and the timing of a fluctuation of a predetermined or more in temperature. Thus, the generation unit 41 may be instructed to update the prediction model. For example, when the time affected by the typhoon is predicted in advance by a weather forecast or the like, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at the time predicted to be affected by the typhoon. May be. Moreover, the update instruction | indication part 44 may instruct | indicate the update of a prediction model with respect to the production | generation part 41, when it is notified that the power failure generate | occur | produced from the other apparatus. In addition, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at a timing when the amount of solar radiation has changed by a predetermined amount or a predetermined ratio as compared to a predetermined time (for example, 30 minutes). May be. The predetermined amount or the predetermined ratio is set to a value that is regarded as a sudden change in the amount of sunlight. Note that the predetermined amount or the predetermined ratio may be determined for each time. Also, the update instruction unit 44 generates the generation unit 41 at a timing when a change of a predetermined amount or a predetermined ratio or more has occurred in the difference between the average temperature of the day and the temperature of the time, compared to a certain time before (for example, 30 minutes). May be instructed to update the prediction model. The predetermined amount or the predetermined ratio is set to a value that is regarded as a sudden change in temperature. Note that the predetermined amount or the predetermined ratio may be determined for each time.

  As described above, according to the prediction method according to the first embodiment, if it is a demand forecast for a short time ahead targeted by the current day correction forecast, it is possible to predict with high accuracy by using the actual value at the forecast time. It became clear. Therefore, the prediction method according to the first embodiment can accurately predict the power demand for the power company. In addition, by updating the prediction model at any one of the timing affected by the typhoon, the timing when a power failure occurs, the timing when fluctuations in the solar radiation amount exceed a predetermined level, or the timing when fluctuations beyond the predetermined level occur in the temperature , It is possible to suppress the occurrence of large deviations in prediction.

[Process flow]
A flow of prediction processing in which the prediction apparatus 10 according to the first embodiment performs prediction will be described. FIG. 16 is a flowchart illustrating an exemplary procedure of a prediction process according to the first embodiment. This prediction process is executed at a predetermined timing, for example, a timing when the prediction device 10 is activated and a predetermined initial process is completed, or a timing when a process start instruction is received.

  As shown in FIG. 16, the update instruction unit 44 determines whether or not the prediction execution time of any of 5:30, 8:30, 10:30, and 14:30 at which the current day correction prediction is performed has been reached ( S10). When it is not the predicted execution time (No in S10), the process proceeds to S10 again.

  On the other hand, when the predicted execution time is reached (Yes at S10), the update instruction unit 44 instructs the generation unit 41 to update the prediction model (S11).

  The production | generation part 41 produces | generates the prediction model which estimates the electric power demand of the time of prediction object using the actual data 30 of the electric power demand obtained by the present time, temperature, and the amount of sunlight (S12). The prediction unit 42 uses the generated prediction model to predict the net demand at the prediction target time from the latest power demand, temperature, and amount of sunshine at the present time (S13). The output control unit 43 causes the display unit 20 to display the information on the net demand at the prediction target time calculated by the prediction unit 42 (S14).

  The update instruction unit 44 determines whether or not an instruction to end a predetermined process has been received (S15). If an instruction to end the process has not been received (No at S15), the process proceeds to S10 described above. On the other hand, if an instruction to end the process is received (Yes at S15), the process ends.

[effect]
As described above, the prediction device 10 according to the present embodiment predicts the power demand at a predetermined prediction target time by using the power demand, temperature, and sunshine amount actual data 30 obtained up to the time when the prediction is performed. Generate a model. The prediction device 10 uses the generated prediction model to predict the power demand at the prediction target time from the power demand at the time of prediction execution, the temperature, and the amount of sunlight. The prediction device 10 updates the prediction model at a predetermined timing. Thereby, the prediction apparatus 10 can predict the electric power demand (net demand) with respect to an electric power company accurately.

  Moreover, the prediction apparatus 10 according to the present embodiment updates the prediction model at a predetermined time when the power demand tendency changes during the day. Thereby, since the prediction model is updated at the timing when the tendency of power demand changes during the day, the prediction device 10 can predict power demand with high accuracy.

  In addition, the prediction device 10 according to the present embodiment is any one of the timing affected by the typhoon, the timing when the power failure occurs, the timing when the solar radiation amount varies more than a predetermined amount, or the timing when the temperature variation occurs more than a predetermined amount. The prediction model is updated at the timing. As a result, the prediction model is updated at a timing when the prediction is likely to deviate greatly, so that the power demand can be accurately predicted.

  Moreover, the prediction apparatus 10 according to the present embodiment instructs the update of the prediction model at a timing when the electric power company reviews the operation plan of the generator. Thereby, since the prediction model is updated at the timing when the power company reviews the generator operation plan, the power company can use accurate power demand when the power company reviews the generator operation plan.

  Further, the prediction device 10 according to the present embodiment updates the prediction model at a timing when the difference between the predicted value of the predicted demand and the actual value of the demand exceeds a certain value. Thereby, since a prediction model is updated when prediction accuracy falls, electric power demand can be predicted accurately.

  Next, Example 2 will be described. The functional configuration of the prediction apparatus according to the second embodiment is the same as that of the first embodiment shown in FIG.

  Incidentally, as shown in FIG. 5, the prediction method according to the first embodiment has high accuracy in prediction immediately after the prediction execution time, and the accuracy deteriorates as time elapses from the prediction execution time. For this reason, it is considered that the prediction accuracy is improved by increasing the frequency of the prediction execution. In addition, by shortening the update frequency, since the actual value of the solar radiation amount can be predicted after 6:00 when the actual value of the solar radiation amount on the day is obtained, it can be expected that the morning prediction error will be reduced. .

  Here, as in the prior art, for example, when forecast values of weather are used for forecasting power demand, the update frequency of weather forecasts is limited, and even if the update frequency of forecast of power demand is increased, the latest Information such as weather data cannot always be reflected in the forecast.

  On the other hand, since the prediction method according to the first embodiment uses only the actual value obtained up to the time when the prediction is performed, it is possible to always perform the prediction reflecting the latest actual value by increasing the update frequency of the prediction. For example, the temperature and global solar radiation used in the prediction equations (1) and (2) can be measured in-house, and the latest data can be obtained almost in real time.

  Therefore, the update instruction unit 44 instructs the generation unit 41 to update the prediction model at regular time intervals. For example, assuming that the update interval of the prediction model is 30 minutes, the power demand ahead of 1 hour 30 minutes is predicted by the prediction model. In this case, the update instruction unit 44 instructs the generation unit 41 to update the prediction model every 30 minutes.

In response to an instruction to update the prediction model from the update instructing unit 44, the generating unit 41 predicts the power demand after a predetermined time every 30 minutes using actual power demand data obtained up to that time. Generate a predictive model. For example, the generation unit 41 uses the performance data for the past two years obtained up to the instructed time point, and uses the prediction formula (2) described above to predict the power demand for 1 hour 30 minutes ahead (a determining the value of ₁ ~a _9).

The prediction unit 42 uses the prediction model generated by the generation unit 41 every 30 minutes to predict the power demand after a predetermined time from the power demand, temperature, and amount of sunshine at that time. For example, the prediction unit 42 uses the prediction formula (2) in which the values of the coefficients (a _{1 to} a ₉ ) determined by the generation unit 41 are set, and calculates the power demand, the temperature, and the amount of sunshine at that time for one hour. Predict power demand 30 minutes ahead.

[Prediction example]
Next, an example of prediction by the prediction device 10 to which the above-described prediction method is applied will be described. In the following, an example will be described in which the prediction accuracy is verified using actual power demand data measured at intervals of 30 minutes from April 1, 2009 to February 28, 2015. Since the prediction formulas (1) and (2) require the past two years of actual data to determine the coefficient, here, from April 1, 2012 when the past two years of actual data are available, 2015 On the day of correction correction until February 28, the prediction accuracy was verified.

  FIG. 17 is a diagram illustrating an example of a verification result of prediction accuracy by the prediction method according to the second embodiment. As shown in FIG. 17, the prediction error for each time is generally smaller than that in FIG. In the prediction method according to the first embodiment, the predicted value of power demand varies from 30 minutes ahead of the prediction execution time to 4 hours 30 minutes ahead. However, in the prediction method according to the second embodiment, It is thought that this is because all the predicted values are only predicted values one hour and thirty minutes ahead. In the prediction method according to the second embodiment, as compared with the prediction method according to the first embodiment, the error is larger in the time zone in which the prediction is performed from 30 minutes ahead to 1 hour 30 minutes ahead. On the other hand, in the prediction method according to the second embodiment, the error is reduced in the time zone in which the time ahead of 1 hour 30 minutes is predicted. In the prediction method according to the second embodiment, the prediction error is constant regardless of time. In the second embodiment, in particular, the prediction accuracy is improved in a time zone in which the time has elapsed since the prediction execution time of the first embodiment and the prediction error is large. In the prediction method according to the first embodiment, there is a time zone in which the prediction error is 2.5% when the prediction model is updated at the timing of the current day correction prediction. On the other hand, in the prediction method according to the second embodiment, when the demand ahead for 1 hour 30 minutes is predicted at 30 minute intervals, the error hardly exceeds 2%. Further, in the prediction method according to the second embodiment, since the prediction after the sunrise reflects the actual value of the amount of solar radiation at the prediction time, the morning prediction error is also reduced. Looking at the difference in prediction error for each year, even when prediction is performed at 30-minute intervals, the prediction error for 2012 is the smallest, and the prediction error increases each year.

  Even in the prediction method according to the second embodiment, a case where the prediction error exceeds 10% is regarded as a great difference. FIG. 18 is a diagram showing the number of days of occurrence for each month when the prediction is significantly different at any time. As shown in FIG. 18, the number of days of outliers is reduced compared to FIG. 6. The reason why the frequency of outliers has decreased is that increasing the frequency of updating forecasts increases the possibility of responding to sudden changes in demand in a short time.

  Here, the result is compared with the day when the prediction of the power demand deviates greatly by the prediction method according to the first embodiment. FIG. 19 is a diagram illustrating the change in the actual value and the predicted value of the power demand on the day when the prediction deviates greatly due to a power failure due to typhoon damage. FIG. 19 shows the result of prediction for the same day as in FIG. In addition, in FIG. 19, it has shown by the value normalized as the daily maximum electric power of the object day is set to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 19, although the prediction error at each time is not small, the prediction value is corrected to correspond to a different demand pattern by updating the prediction at 30-minute intervals compared to FIG. Has been.

  Even if the forecast is out of focus due to sudden changes in temperature or solar radiation, the forecast update interval is set to 30 minutes, and the forecast will be revised to reflect changes in the latest temperature and the actual amount of sunshine. I can confirm that.

  FIG. 20 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when the sudden change in the amount of sunshine occurs. FIG. 20 shows the result of prediction for the same day as in FIG. In FIG. 20, the electric power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 20, even when the amount of solar radiation changes suddenly, the prediction is adapted to changes in demand due to changes in the amount of solar radiation by updating the prediction at 30-minute intervals compared to FIG. Can be confirmed.

  FIG. 21 is a diagram illustrating changes in the actual value and the predicted value of the power demand on the day when a sudden change in temperature occurs. FIG. 21 shows the result of prediction for the same day as in FIG. In FIG. 21, the electric power demand is normalized with a predetermined reference value of 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 21, even when the temperature suddenly changes, it is confirmed that the prediction is adapted to the change in demand due to the change in temperature by updating the prediction at intervals of 30 minutes as compared with FIG. it can.

  FIG. 22 is a diagram showing changes in the actual value and the predicted value of the power demand on the day when the temperature and the amount of solar radiation suddenly change. FIG. 22 shows the result of prediction for the same day as in FIG. In FIG. 22, the power demand is normalized by setting the predetermined reference value to 1.0. The solid line represents the predicted value. The dotted line represents the actual value. As shown in FIG. 22, even when the temperature and the amount of solar radiation change suddenly, compared with FIG. 13, by updating the prediction at intervals of 30 minutes, It can be confirmed that

  However, shortening the prediction interval may lead to an increase in prediction error. In the case of a power outage due to equipment failure, if the current forecast is corrected on the same day, the power outage was restored before the forecast was updated, so the prediction error decreased after the power outage was restored and demand returned to the normal level. Yes. On the other hand, when prediction is performed at 30-minute intervals, the prediction of the power demand at 17:30 is lower than usual because the prediction at the time when the power failure occurs is updated to reflect the actual demand value. It is predicted and the error is large.

  Therefore, for example, the update instruction unit 44 predicts the generation unit 41 at a timing when the difference between the predicted demand value predicted by the prediction unit 42 and the actual demand value exceeds a certain value (for example, 2%). A model update may be instructed. For example, the update instruction unit 44 is the time when the difference between the predicted value of the demand predicted by the prediction unit 42 and the actual value of the demand exceeds a certain value (for example, 2%) and the difference exceeds the certain value. For subsequent predicted times, an update of the prediction model may be instructed, and prediction of power demand at the predicted time may be executed again. The constant value may be determined for each time. Further, the fixed value may be changeable from the outside by an administrator or the like.

[Process flow]
A flow of prediction processing in which the prediction device 10 according to the second embodiment performs prediction will be described. FIG. 23 is a flowchart illustrating an exemplary procedure of a prediction process according to the second embodiment. This prediction process is executed at a predetermined timing, for example, a timing when the prediction device 10 is activated and a predetermined initial process is completed, or a timing when a process start instruction is received.

  As shown in FIG. 23, the update instruction unit 44 determines whether or not a predetermined time has elapsed since the previous prediction (S50). If the predetermined time has not elapsed (No at S50), the process proceeds to S10 again. In the case of the first prediction, the update instruction unit 44 determines that a certain time has passed since the previous prediction.

  On the other hand, when the fixed time has elapsed (Yes at S50), the update instruction unit 44 instructs the generation unit 41 to update the prediction model (S51).

  The production | generation part 41 produces | generates the prediction model which estimates the electric power demand after predetermined time using the performance data 30 of the electric power demand obtained by the present time, temperature, and the amount of sunlight (S52). The prediction unit 42 uses the generated prediction model to predict the net demand after a predetermined time from the current power demand, temperature, and amount of sunlight (S53). The output control unit 43 causes the display unit 20 to display the information on the net demand at the prediction target time calculated by the prediction unit 42 (S54).

  The update instruction unit 44 determines whether or not an instruction to end a predetermined process has been received (S55). If an instruction to end the process has not been received (No at S55), the process proceeds to S50 described above. On the other hand, if an instruction to end the process is received (Yes at S55), the process ends.

[effect]
As described above, the prediction device 10 according to the present embodiment generates a prediction model that predicts the power demand after a predetermined time by using the power demand actual data obtained up to the time point for every predetermined time. . The prediction device 10 predicts the power demand after a predetermined time from the power demand, the temperature, and the amount of sunlight at a certain time using the generated prediction model at regular time intervals. Thereby, the prediction apparatus 10 can predict the electric power demand (net demand) with respect to an electric power company accurately.

  Further, the prediction device 10 according to the present embodiment updates the prediction model at a timing when the difference between the predicted value of the predicted demand and the actual value of the demand exceeds a certain value. Thereby, since a prediction model is updated when prediction accuracy falls, electric power demand can be predicted accurately.

  Although the embodiments related to the disclosed apparatus have been described so far, the disclosed technology may be implemented in various different forms other than the above-described embodiments. Therefore, another embodiment included in the present invention will be described below.

  For example, in the above-described embodiment, the case where the prediction formulas (1) and (2) are used as the prediction model for predicting the power demand from the actual data of the power demand, the temperature, and the amount of sunlight has been described. It is not limited to this. For example, other prediction formulas may be used as long as the power demand can be predicted from the actual data of power demand, temperature, and amount of sunlight.

  Moreover, although the said Example demonstrated the case where a prediction model was produced | generated using the performance data 30 of the electric power demand, temperature, and sunshine amount obtained by the prediction implementation time, the apparatus of an indication is not limited to this. For example, when the error of the power demand due to the influence of the temperature and the amount of sunlight is allowed, the prediction model may be generated using only the actual data of the power demand obtained up to the prediction execution time. For example, the prediction formulas may be used by excluding the terms “cooling / heating effect variable” and “sunlight effect variable” from the prediction formulas (1) and (2). Since the electric power company can obtain actual power demand data in-house, the electric power company can predict the power demand using only the data obtained in-house. Further, as described above, if the demand forecast is a short time ahead, it is possible to predict with high accuracy by using the actual value at the time of prediction. For this reason, even if the prediction model is generated using only the actual power demand data obtained up to the time when the prediction is performed, the above-described great difference can be obtained if it is a short time of about several hours (for example, 4 hours). Unless it is a causal case, power demand can be predicted with an acceptable prediction accuracy. The allowable time for allowing the prediction of power demand in the prediction model generated only from the actual data 30 varies depending on the allowable prediction accuracy. When the allowable prediction accuracy is high, the allowable time is as short as about 1 hour, for example. When the allowable prediction accuracy is low, the allowable time is as long as about 6 hours, for example.

  In the first embodiment, the prediction model is updated to 5:30, 8:30, 10:30, and 14:30, which are the prediction execution times in the current day correction prediction. In the second embodiment, a predetermined time (for example, However, the disclosed apparatus is not limited to this. For example, the update instruction unit 44 may determine an update interval for updating the prediction model using the record data 30. For example, the update instruction unit 44 may determine the update interval by the following steps 100 to 104.

  In step 100, a target error E is set. The target error E may be specified by a ratio (for example, 2%) or a value (for example, 10 kW). In step 100, time t = ts (ts is the prediction start time) is set. In step 100, an update interval Δt = Δt0 (Δt0 is an initial value of the update interval) is set.

  In step 101, a prediction error in the case where the power demand at time t + Δt is predicted from the actual value at time t is calculated by using past actual data for a certain period.

  In step 102, if prediction error ≦ target error E, the process proceeds to step 103. On the other hand, in step 102, when prediction error> target error E, the update interval Δt is shortened (for example, Δt = Δt−ε: ε is a value set in advance), and the process returns to step 101.

  In step 103, when the time t exceeds the predicted date (time t ≧ 24: 00 (24:00)), the process proceeds to step 104. On the other hand, in step 103, when the time t is within the predicted date (time t <24:00), the process returns to step 101 as time t = t + Δt.

  In step 104, the update interval Δt is set to a predicted update time for achieving the target accuracy.

  Further, for example, the update instruction unit 44 may determine an update time for updating the prediction model using the record data 30. For example, the update instruction unit 44 may determine the update time by the following steps 150 to 154.

In step 150, a target error E is set. The target error E may be specified by a ratio (for example, 2%) or a value (for example, 10 kW). In step 150, the variable n is initialized to zero (n = 0). In step 150, time t ₀ = ts (ts is the prediction start time) is set. In step 150, Δt0 (initial value of the update interval) is set.

In step 151, using the actual data of the past predetermined period, it calculates a prediction error in the case where the predicted using actual values of the time t _{n + 1} the power demand time t _n.

In step 152, if prediction error ≦ target error E, time t _n is advanced to the future (t _n = t _n + ε), and the process returns to step 151. On the other hand, in step 152, when the prediction error> the target error E, the time t _n is returned to the past (t _n = t _n −ε), and the process proceeds to step 153.

In step 153, when the time t _n exceeds the predicted date (time t _n ≧ 24: 00 (24:00)), the process proceeds to step 154. On the other hand, in step 153, when the time t _n is within the predicted date (time t _n <24:00), the process returns to step 151 as time t _{n + 1} = time t _n + Δt0.

In step 154, sets the time t _n to the update time of the prediction, each achieving the target accuracy.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific state of distribution / integration of each device is not limited to the one shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. For example, the processing units of the acquisition unit 40, the generation unit 41, the prediction unit 42, the output control unit 43, and the update instruction unit 44 may be appropriately integrated. Further, the processing of each processing unit may be appropriately separated into a plurality of processing units. Further, all or any part of each processing function performed in each processing unit can be realized by a CPU and a program analyzed and executed by the CPU, or can be realized as hardware by wired logic. .

[Prediction program]
The various processes described in the above embodiments can also be realized by executing a program prepared in advance on a computer system such as a personal computer or a workstation. Therefore, in the following, an example of a computer system that executes a program having the same function as in the above embodiment will be described. FIG. 24 is a diagram illustrating a computer that executes a prediction program.

  As shown in FIG. 24, the computer 300 includes a central processing unit (CPU) 310, a hard disk drive (HDD) 320, and a random access memory (RAM) 340. These units 300 to 340 are connected via a bus 400.

  The HDD 320 stores in advance a prediction program 320 a that performs the same functions as the acquisition unit 40, generation unit 41, prediction unit 42, output control unit 43, and update instruction unit 44. Note that the prediction program 320a may be separated as appropriate.

  The HDD 320 stores various information. For example, the HDD 320 stores various data used for prediction, such as the above-described performance data 30.

  And CPU310 reads the prediction program 320a from HDD320, and performs the operation | movement similar to each process part of an Example. That is, the prediction program 320a performs the same operations as the acquisition unit 40, the generation unit 41, the prediction unit 42, the output control unit 43, and the update instruction unit 44.

  Note that the above-described prediction program 320a is not necessarily stored in the HDD 320 from the beginning.

  For example, the program is stored in a “portable physical medium” such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, or an IC card inserted into the computer 300. Then, the computer 300 may read and execute the program from these.

  Furthermore, the program is stored in “another computer (or server)” connected to the computer 300 via a public line, the Internet, a LAN, a WAN, or the like. Then, the computer 300 may read and execute the program from these.

DESCRIPTION OF SYMBOLS 10 Prediction apparatus 20 Display part 21 Input part 22 Storage part 23 Control part 30 Performance data 40 Acquisition part 41 Generation part 42 Prediction part 43 Output control part 44 Update instruction part

Claims (9)

A generation unit that generates a prediction model for predicting the power demand at a predetermined prediction target time using the actual power demand data obtained up to the prediction execution time point;
A prediction unit that predicts the power demand at the prediction target time from the power demand at the time of the prediction using the prediction model generated by the generation unit,
An update instruction unit that instructs the generation unit to update the prediction model at a predetermined timing;
The prediction apparatus characterized by having. The generation unit generates a prediction model that predicts the power demand at the prediction target time using the power demand, temperature, and sunshine amount data obtained up to the prediction execution time point,
The prediction unit predicts the power demand at the prediction target time from the power demand, temperature, and amount of sunshine at the time of the prediction using the prediction model generated by the generation unit. The prediction device described. The prediction apparatus according to claim 1, wherein the update instruction unit instructs the generation unit to update the prediction model at a predetermined time when a power demand tendency changes during a day. . The update instructing unit is the generation unit at any one of a timing affected by a typhoon, a timing when a power failure occurs, a timing when a variation of the solar radiation amount exceeds a predetermined level, or a timing when a variation of the temperature level exceeds a predetermined level. The prediction apparatus according to claim 1, wherein the prediction model is instructed to be updated. The said update instruction | indication part instruct | indicates the update of the said prediction model with respect to the said production | generation part at the timing in which an electric power company reviews the driving | operation plan of a generator. The Claim 1 characterized by the above-mentioned. Prediction device. The update instruction unit instructs the generation unit to update the prediction model at a timing when a difference between a predicted value of demand predicted by the prediction unit and an actual value of demand exceeds a certain value. The prediction apparatus according to any one of claims 1 to 5. The update instruction unit instructs the generation unit to update the prediction model at regular time intervals,
In response to an instruction to update the prediction model from the update instruction unit, the generation unit calculates the power demand after a predetermined time by using the actual power demand data obtained up to that point in time. Generate a predictive model to predict,
The prediction unit predicts the power demand after the predetermined time from the power demand, the temperature, and the amount of sunlight at the time point using the prediction model generated by the generation unit at every predetermined time. The prediction apparatus according to claim 1 or 2. Using the actual power demand data obtained up to the time of forecasting, generate a forecast model that forecasts power demand at a given forecast target time,
Using the generated prediction model, predict the power demand at the prediction target time from the power demand at the time of the prediction execution,
A prediction method characterized in that a computer executes a process of instructing an update of the prediction model at a predetermined timing. Using the actual power demand data obtained up to the time of forecasting, generate a forecast model that forecasts power demand at a given forecast target time,
Using the generated prediction model, predict the power demand at the prediction target time from the power demand at the time of the prediction execution,
A prediction program that causes a computer to execute a process of instructing updating of the prediction model at a predetermined timing.

Images