JP2016073155A - Demand prediction device, demand prediction method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To add information concerning prediction reliability to a prediction concerning power demand.SOLUTION: The demand prediction device for predicting power demand includes a demand calculation part for calculating a plurality of prediction values concerning power demand of a power load at a specified date and time on the basis of a prediction formula representing a relation between power demand at the power load and meteorological information at a location of the power load, and a plurality of meteorological prediction values at the location of the power load and a specified date and time. The demand prediction device further includes a reliability determination part for determining reliability with respect to the predicted power demand on the basis of a distribution in the collection of the plurality of prediction values.SELECTED DRAWING: Figure 4

Description

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

  It is known that power demand is related to weather information such as temperature and humidity. Therefore, power demand prediction is closely related to weather forecast. As a method for predicting power demand, for example, a weather forecast model is created based on GPV (Grid Point Value) data distributed from the Japan Meteorological Agency and observed weather information, and past power demand results are referred to There is a method for predicting power demand corresponding to the predicted temperature and humidity (Patent Document 1).

JP 2003-180032 A

  In general, weather forecast information such as GPV data includes one forecast value for one type of weather condition at a certain date and time, and such forecast information has an error due to incompleteness of the underlying forecast model. Because it is inherent, the reliability of forecast information is not clear. If the error in the forecast information is large, the error is accumulated in the process of calculating the demand forecast value, greatly affecting the accuracy of the demand forecast, and may possibly impair the reliability of the forecast.

  The main present invention that solves the above-described problem is a demand prediction device that predicts power demand, a prediction formula that shows a relationship between power demand in a power load and weather information at the position of the power load, A demand calculation unit that calculates a plurality of predicted values related to the power demand of the power load at the specified date and time based on a plurality of weather forecast values for the specified date and time at the position; and a set of the plurality of predicted values And a reliability determination unit that determines the reliability of the predicted power demand based on the distribution in FIG.

  Other features of the present invention will become apparent from the accompanying drawings and the description of this specification.

  According to the present invention, it is possible to add information related to the reliability of prediction to the prediction related to power demand.

It is the schematic which shows the electric power grid | system to which the prediction apparatus in embodiment of this invention is applied. It is the schematic which shows the area | region and mesh to which the prediction apparatus in embodiment of this invention is applied. It is a block diagram which shows the function of a prediction apparatus in embodiment of this invention. It is a flowchart which shows operation | movement of the prediction apparatus in embodiment of this invention. It is a figure which shows an example of the weather observation data in embodiment of this invention. It is a figure which shows an example of the demand performance data in embodiment of this invention. It is a figure which shows an example of the weather forecast data in embodiment of this invention. It is a figure which shows an example of the demand prediction data in embodiment of this invention. It is a figure which shows an example of the total demand prediction data in embodiment of this invention. (A) The figure which shows an example of the time change of the actual value of a total demand, and several predicted values in embodiment of this invention, (b) The figure which shows the frequency distribution of the predicted value of the total demand at the time t1, and (c) It is a figure which shows the frequency distribution of the predicted value of the total demand at the time t2.

  At least the following matters will become apparent from the description of this specification and the accompanying drawings.

=== Application Scene of Prediction Device ===
Hereinafter, with reference to FIG. 1 and FIG. 2, a typical scene where the prediction apparatus according to the present embodiment is applied will be described. FIG. 1 schematically shows a power system to which the prediction device according to the present embodiment is applied, and FIG. 2 schematically shows regions and meshes to which the prediction device is applied.

  The power system 200 is a power system for supplying power to the power load. As shown in FIG. 1, the power load R1 connected in parallel to the commercial power supply G and the power line connected to the commercial power supply G, respectively. -R4 is included. The power loads R1 to R4 are, for example, power devices (such as air conditioners) in general consumers, and for the sake of simplification of description, a predetermined area A (for example, the Chugoku region of Japan) is partitioned as shown in FIG. It is assumed that each of the formed meshes 1 to 4 is installed, and the prediction device predicts the total demand in such a predetermined area A. Note that meshes 1 to 4 correspond to a plurality of areas and correspond to ensemble forecasts described later. Moreover, it is not necessary to install only one power load on each mesh, and a plurality of power loads may be provided. For example, when a plurality of power loads are installed on the mesh 1, the power load R1 is displayed on behalf of such a plurality of power loads.

  As shown in FIG. 1, measuring devices M <b> 1 to M <b> 4 and weather observation devices W <b> 1 to W <b> 4 are installed in each of the meshes 1 to 4. The measuring devices M1 to M4 are devices that measure the powers P1 to P4 supplied from the power system 200 to the power loads R1 to R4, respectively, and output the measurement signals SM1 to SM4 to the prediction device 100. The weather observation devices W1 to W4 observe the weather conditions (for example, temperature, humidity, precipitation, and solar radiation) in the meshes 1 to 4 at predetermined time intervals (for example, every hour), and observe signals SW1 to SW4. Is output to the prediction device 100.

  The weather forecast device WF is a device that outputs a forecast signal SWF indicating ensemble forecast information, and is installed in, for example, the Japan Meteorological Agency. The ensemble forecast information includes a plurality of sets of forecast information indicating the temperature, humidity, precipitation, wind speed, solar radiation, and cloud cover in each of the meshes 1 to 4.

  As will be described in detail later, the prediction device 100 receives the measurement signals SM1 to SM4 from the measurement devices M1 to M4, the observation signals SW1 to SW4 from the weather observation devices W1 to W4, and the prediction signal SWF from the weather prediction device WF, respectively. It is a device that receives and predicts power demand in meshes 1 to 4 and a predetermined area A based on information included in these signals.

  In addition, in order to supply electric power to electric power load R1-R4, the distributed power sources, such as a solar power generation device (PV), may be installed in the meshes 1-4. In this case, the electric power that could not be covered by the distributed power supply is supplied from the electric power system 200 to the electric power loads R1 to R4.

  The prediction device 100 is, for example, a computer including a CPU, a memory, and an auxiliary storage device, and the functions described below that the prediction device has are executed by a program that can be executed by the computer.

=== Prediction device ===
Hereinafter, the prediction apparatus according to the present embodiment will be described with reference to FIGS. 3 and 5 to 9. FIG. 3 is a block diagram illustrating the function of the prediction apparatus according to the present embodiment, FIG. 5 is an example of weather observation data, FIG. 6 is an example of demand performance data, FIG. 7 is an example of weather forecast data, 8 shows an example of demand forecast data, and FIG. 9 shows an example of total demand forecast data.

<Outline of prediction device>
The prediction device 100 predicts the power demand for each mesh 1 to 4. Specifically, the prediction device 100 receives, for example, the measurement signal SM1 from the measurement device M1, the observation signal SW1 from the weather observation device W1, and the prediction signal SWF from the weather prediction device WF in order to predict the power demand in the mesh 1, for example. A plurality of predicted values related to the power (power demand) supplied from the power system 200 to the power load R1 are calculated based on information received and included in these signals. Similarly for the meshes 2 to 3, the prediction device 100 calculates a plurality of predicted values related to the power (power demand) supplied from the power system 200 to the power loads R <b> 2 to R <b> 4. This corresponds to the function of the first calculation unit.

  Further, the prediction device 100 predicts the total demand in the predetermined area A. Specifically, the prediction device 100 uses the power (total demand) supplied from the power system 200 to the power loads R1 to R4 in the predetermined area A based on a plurality of predicted values calculated for each of the meshes 1 to 4. ) To calculate a plurality of predicted values. This corresponds to the function of the second calculation unit.

  The prediction apparatus 100 further determines the reliability of the prediction based on the calculated distribution of the plurality of prediction values. This corresponds to the function of the reliability determination unit.

  Moreover, the prediction apparatus 100 determines the estimated value of the total demand in the predetermined area A based on the determined reliability. This corresponds to the function of the expected value determination unit.

  As shown in FIG. 3, the prediction device 100 includes a transmission / reception unit 110, a storage unit 120, a prediction formula derivation unit 130, a demand calculation unit 140, and an output unit 150.

<Transmitter / receiver>
The transmission / reception unit 110 performs wired or wireless communication with each of the measurement devices M1 to M4, the weather observation devices W1 to W4, and the weather forecast device WF, and receives the measurement signals SM1 to SM4, the observation signals SW1 to SW4, and the forecast signal SWF. To do. In the present embodiment, each of these signals is output from each device at a predetermined time interval, but may be output in response to a request signal from the transmission / reception unit 110.

<Storage unit>
The storage unit 120 includes a weather observation data storage area 121, a demand performance data storage area 122, a PV introduction data storage area 123, a prediction formula storage area 124, a weather forecast data storage area 125, and a demand prediction data storage area 126.

  The meteorological observation data storage area 121 stores meteorological observation data provided from the meteorological observation apparatuses W1 to W4 at predetermined time intervals for the meshes 1 to 4. In the present embodiment, observation data relating to hourly temperature, humidity, precipitation, and solar radiation, for example, as shown in FIG. 5 is stored for each mesh.

  In the demand record data storage area 122, measurement data (demand record data) regarding the power demand for each predetermined time interval in the meshes 1 to 4 provided from the measurement devices M1 to M4 is stored. In the present embodiment, hourly demand data as shown in FIG. 6, for example, is stored for each mesh.

  In the PV introduction data storage area 123, information (PV introduction data) indicating the introduction amount of the photovoltaic power generation apparatus in the meshes 1 to 4 is stored. The PV introduction data is represented by a numerical value such as 5 MW, for example.

  The prediction formula storage area 124 stores a prediction formula, which will be described later, indicating the relationship between weather conditions and power demand, which is derived for each mesh 1 to 4 by the prediction formula deriving unit 130.

  The weather forecast data storage area 125 stores weather forecast data for each mesh 1 to 4 provided from the weather forecast device WF. The weather forecast data in the present embodiment is ensemble forecast data, and is provided from the weather forecast device WF at intervals of 6 hours. Among several types of ensemble forecasts, weekly ensemble numerical forecasts covering Japan are used, and the weather conditions indicated by 27 ensemble members with initial values of 00 UTC (Coordinated Universal Time) and 12 UTC are 264 from the current time. It is forecasted until the time ahead. The weekly ensemble numerical forecast includes, for example, forecast values related to wind speed in the east-west and north-south directions on the ground, temperature, relative humidity, accumulated precipitation, total cloud cover, sea level correction pressure, and ground pressure. For example, as shown in FIG. 7, 27 sets of forecast data relating to temperature, humidity, precipitation, and solar radiation (cloud cover) for each predetermined date and time are stored in the storage area 125. Of course, other information such as wind speed and atmospheric pressure may be stored in the weather forecast data storage area 125.

  The demand prediction data storage area 126 stores a plurality of prediction values (area prediction values) related to the power demand for each mesh 1 to 4 calculated by the first calculation unit of the demand calculation unit 140. In this embodiment, corresponding to the fact that there are 27 sets of weather forecast data used for prediction of power demand for each mesh 1 to 4 per predetermined date and time as in the example of FIG. 7, the area prediction values are illustrated in FIG. 8. There are 27 per predetermined date and time. The calculation procedure of the predicted value of the power demand for each mesh 1 to 4 will be described later.

  The demand prediction data storage area 126 stores a plurality of prediction values (region prediction values) related to the total demand in the predetermined area A calculated by the second calculation unit of the demand calculation unit 140. Although the calculation procedure of the predicted value of the total demand will be described later, in the present embodiment, as the four areas are included in the predetermined area A, the predicted value of the total demand is a predetermined date and time as illustrated in FIG. There are 531,441 (27 to the 4th power).

<Prediction formula deriving unit>
The prediction formula deriving unit 130 derives a prediction formula for each mesh 1 to 4 based on the meteorological observation data stored in the meteorological observation data storage area 121 and the demand record data stored in the demand record data storage area 122. As will be described later, the prediction formula is a formula for predicting the power demand in each of the meshes 1 to 4 and is given as a linear regression equation obtained by multiple regression analysis in the present embodiment.

<Demand calculation part>
The demand calculation unit 140 is a part that calculates a predicted value of power demand for each mesh 1 to 4 and a predicted value of total demand in the predetermined area A, and includes a first calculation unit 141, a second calculation unit 142, and reliability determination. Part 143 and an expected value determination part 144. The first calculation unit 141 includes a plurality of prediction values related to power demand in each of the meshes 1 to 4 based on the prediction formula generated by the prediction formula deriving unit 130 and the weather forecast data stored in the weather forecast data storage area 125. Is calculated. The second calculation unit 142 calculates a plurality of prediction values related to the total demand in the predetermined area A based on the plurality of prediction values related to the power demands of the meshes 1 to 4 calculated by the first calculation unit 141. The reliability determination unit 143 determines the reliability of prediction based on the distribution of a plurality of prediction values related to the total demand calculated by the second calculation unit 142. The distribution of the predicted values may be indicated by, for example, an arithmetic mean and a standard deviation, or may be given by a frequency distribution diagram as shown in FIGS. 10 (b) and 10 (c). Then, the expected value determining unit 144 determines the estimated value of the total demand in the predetermined area A based on the prediction reliability determined by the reliability determining unit 143. In addition, the calculation procedure of the predicted value regarding the power demand in the meshes 1 to 4 and the total demand in the predetermined area A, the determination procedure of the reliability of the prediction, and the determination procedure of the expected value of the total demand will be described later.

  Note that the reliability determination unit 143 may obtain the reliability of prediction for each mesh based on the distribution of a plurality of prediction values related to the power demand for each mesh 1 to 4, and the prospective value determination unit 144 may Based on the obtained reliability for each mesh, an expected value of power demand for each mesh may be determined.

<Output unit, etc.>
The output unit 140 is a part for displaying information input to the prediction device 100 or displaying information output from the prediction device 100, and is, for example, a monitor or a printer. The prediction device 100 may include a keyboard for inputting information to the prediction device 100.

=== Forecasting procedure for demand ===
With reference to FIG. 4, FIG. 10, etc., the forecasting procedure of the demand performed with the prediction apparatus 100 in this embodiment is demonstrated. FIG. 4 is a flowchart showing the operation of the prediction device 100. FIG. 10 shows (a) an example of the temporal change in the actual value of the total demand and a plurality of predicted values, (b) the frequency distribution of the predicted value of the total demand at a certain time t1, and (c) after the time t1 (future). The frequency distribution of the predicted value of the total demand at time t2 is shown.

<Outline of prediction procedure>
In this embodiment, for convenience of explanation, as shown in FIGS. 5 and 6, for example, weather observation data and demand record data are provided from the weather observation devices W1 to W4 and the measurement devices M1 to M4, respectively, every hour. It is assumed that at least the past one year of data is stored in the observation data storage area 121 and the demand result data storage area 122. Further, as illustrated in FIG. 7, prediction is made when ensemble forecast data including a weather forecast after the date and time (future) is provided from the weather forecast device WF, with 0:00 on August 1, 2014 as the current time. The procedure in which the device 100 predicts the power demand is shown.

First, the power demand prediction procedure will be schematically described. As shown in FIG. 4, in step S1, a prediction formula for power demand is generated for each mesh 1 to 4, and in step S2, for each mesh 1 to 4, A weather forecast value is substituted into the generated forecast formula to calculate a plurality of forecast values related to power demand. In step S3, a plurality of predicted values related to the total demand in the predetermined area A are obtained based on the calculated predicted values for each of the meshes 1-4. In step S4, a distribution of a plurality of predicted values related to the total demand is obtained, and the reliability of the predicted total demand is determined based on the distribution. In step S5, the total demand is estimated based on the determined reliability. Determine the value. Note that these procedures may be executed, for example, every time new weather forecast data is provided from the weather forecast device WF (for example, every 6 hours).
Hereinafter, steps S1 to S5 will be described in detail.

<Generation of prediction formula>
The generation of the prediction formula in step S1 of FIG. 4 is executed by the prediction formula deriving unit 130.

  The prediction formula deriving unit 130 is based on the weather observation value in the past predetermined period (for example, the past one year from the present), the PV introduction data, and the actual demand value in the past predetermined period for each mesh 1 to 4. Then, a prediction formula representing the relationship between these values is derived.

  Specifically, in order to obtain a prediction formula for each mesh, first, the prediction formula deriving unit 130 reads weather observation data (for example, temperature, humidity, precipitation amount, etc.) of the mesh 1 from the weather observation data storage area 121 in the past predetermined period. The amount of solar radiation (cloud amount) (see FIG. 5) is acquired, and the demand result data of the past predetermined period is acquired from the demand result data storage area 122. In addition, the prediction formula deriving unit 130 calls the PV introduction data from the PV introduction data storage area 123.

And the prediction formula derivation | leading-out part 130 performs multiple regression analysis based on the acquired weather observation data, PV introduction data, and a demand actual value, and calculates | requires the prediction formula showing the relationship of these each value. The prediction formula is given by, for example, a linear regression equation represented by the following formula (1), and the prediction formula deriving unit 130 may derive this prediction formula using the least square method.
_{Y = a + b · X 1} + c · X 2 + d · X 3 + e · X 4 · PV (1)
Here, X ₁ is an air temperature, X ₂ is a humidity, X ₃ is a precipitation variable, X ₄ is a solar radiation amount, and Y is a regression variable indicating a demand actual value. Further, a to e are regression parameters and can be determined by the least square method as described above. PV is a constant based on PV introduction data for each of sites 1 to 4. PV = 0 when no photovoltaic power generation apparatus is installed at the corresponding site.

  When the regression parameters a to e of the formula (1) are determined in this way, a prediction formula for power demand in the mesh 1 is determined and stored in the prediction formula storage area 124. Then, the prediction formula deriving unit 130 repeats the above-described procedure until the power demand prediction formulas are derived for all the meshes 2 to 4.

By substituting the weather forecast data into the regression variables X _{1 to} X ₄ of the prediction formula thus derived, the predicted value (Y) of the power demand is obtained for each mesh 1 to 4 in the next step S2. Will be able to. Moreover, by generating a prediction formula for each of the meshes 1 to 4, it becomes possible to predict the power demand in consideration of the situation unique to the meshes 1 to 4, and the prediction accuracy is improved.

  In deriving the prediction formula, time zone, day of the week, holiday / weekday distinction, special day, etc. may be considered. For example, electric power demand generally tends to increase during the daytime during the day and decrease during the nighttime. Even during the same time period, it is known that demand increases or decreases on a specific day of the week, and holiday power demand tends to decrease more than power demand on weekdays. Therefore, for example, it is conceivable to add a term that reflects the increase or decrease in demand due to a time zone or a specific day to the right side of the obtained regression equation (1). Thus, the prediction accuracy of electric power demand improves by generating a prediction formula in consideration of a time zone and a specific day.

<Calculation of predicted power demand for each mesh>
When the prediction formula is derived, in step S2, the forecast value of the future power demand is calculated for each of the meshes 1 to 4 by applying the weather forecast data to the derived prediction formula. Such calculation of the predicted value is executed by the first calculation unit 141 of the demand calculation unit 140.

  Specifically, the first calculation unit 141 first acquires the weather forecast data of the mesh 1 stored in the weather forecast data storage area 125. Such weather forecast data is numerical data based on the ensemble forecast as described above, and weather conditions at a specific date and time in one mesh are given as a set of 27 numerical values. In this embodiment, as illustrated in FIG. 7, 27 sets of predicted values relating to temperature, humidity, precipitation, and solar radiation (cloud cover) every hour after the current time (August 1, 2014, 0:00) Are stored in the weather forecast data storage area 125 to be used for calculation of the predicted value.

  Next, the first calculation unit 141 applies the 27 sets of forecast values at the designated date and time to the prediction formula (1), and obtains a plurality of forecast values related to the power demand of the mesh 1 at this date and time. A plurality of these predicted values are calculated for one date and time of one mesh in accordance with 27 sets of ensemble predicted values.

  After obtaining the predicted value of electricity demand of the mesh 1 at the designated date and time in this way, the designated date and time is as of August 1, 2014 at 1 o'clock, 2 o'clock on the same day, ... The above procedure is repeated by changing only a predetermined time interval (1 hour in the present embodiment) toward the future. By repeating such a procedure, the first calculation unit 141 predicts the power demand in the mesh 1 to several tens of hours ahead. The predicted value for the next several tens of hours related to the power demand in the mesh 1 predicted in this way is stored in the demand prediction data storage area 126. In addition, the prediction result of the electric power demand in the mesh 1 becomes like the data table of FIG. 8, for example.

  And the 1st calculation part 141 repeats the above-mentioned procedure until it calculates the predicted value of an electric power demand about all the meshes 2-4, Thereby, step S2 is complete | finished and it moves to step S3. Note that the predicted value for each mesh 1 to 4 calculated in this way may be output via the output unit 150.

<Calculation of the forecast value of total demand in a given area>
When the predicted value of the future power demand is calculated for each of the meshes 1 to 4, a plurality of predicted values related to the future total demand in the predetermined area A are calculated in step S3. The calculation of the predicted value of the future total demand is executed by the second calculation unit 142 of the demand calculation unit 140.

  In the present embodiment, the predicted value of the future total demand is calculated based on the predicted value of the power demand of the meshes 1 to 4 at the specified date (August 1, 2014, 0:00). In the present embodiment, where there are 27 predicted values for each of the meshes 1 to 4 at the designated date and time, the second calculation unit 142 selects and selects one arbitrary predicted value from each of the meshes 1 to 4. The sum of the predicted values of meshes 1 to 4 is taken as one predicted value of total demand. And the 2nd calculation part 142 performs the same operation | work about all the combinations in the predicted value of the meshes 1-4. Then, in this embodiment, the predetermined area A includes four meshes, and there are 27 predicted values for the designated date and time in each mesh, so that there are 27 4 (531,441) combinations as described above. Become. Therefore, the predicted value of the total demand is calculated by 27 4 to the fourth power (531,441).

  In general, there are n meshes in a predetermined area A, and m sets of weather forecast values are given for a specific date and time for each mesh (thus, m power demand prediction values in each mesh). The predicted value of the total demand calculated by the second calculation unit 142 is m to the nth power per date. In the present embodiment, four meshes are used for simplification of explanation, but when the predetermined area A includes more meshes, the number of predicted total demand values calculated in step S3 increases.

  When the plurality of predicted values of the total demand at the specified date and time are calculated in this way, the second calculation unit 142 sets the specified date and time at 1 o'clock on August 1, 2014, 2 o'clock on the same day, and 0 o'clock on the same month, 2 o'clock. .., And the same calculation is repeated by changing only a predetermined time interval (1 hour in the present embodiment) toward the future. By repeating such calculation, the second calculation unit 142 calculates the total demand in the predetermined area A up to several tens of hours ahead. Then, the predicted value calculated for tens of hours related to the total demand in the predetermined area A is stored in the demand prediction data storage area 126. Moreover, the prediction result of total demand becomes like the data table of FIG. 9, for example. And if the 2nd calculation part 142 calculates total demand to several tens of hours ahead, step S3 will be complete | finished and it will transfer to step S4.

  Thus, by calculating the sum of one predicted value selected from each of the meshes 1 to 4 for all the combinations to be the predicted value of the total demand, there is an error inherent in each predicted value for each mesh. Expect to cancel each other out. Therefore, compared with the case where there is only one predicted value for one date and time, power demand can be predicted in consideration of weather risks, and the reliability of the prediction is improved. This contributes to efficient and stable operation of the power generation plan in the predetermined area A.

  In addition, the graph of the total demand over the fixed period in the future calculated in step S3 is generally shown as in the example of FIG. In other words, the horizontal axis is the time axis and the vertical axis is the total demand, and the actual value and multiple predicted values (however, the predicted values of the same member at different times are connected with a straight line to show the time change for each ensemble member) Is a general tendency that the variation (dispersion degree) of the predicted value increases as time progresses from the current time t0 to t1 and t2 toward the future. This can be understood from the state of the frequency distribution diagram (histogram) of the predicted value of the total demand at the times t1 and t2 in FIG. 10 (a) as shown in FIGS. 10 (b) and 10 (c). In addition, the curve on the left side from the current time t0 in FIG. The graph shown in FIG. 10 may be output to a monitor or the like via the output unit 150.

  Note that the combination of the predicted values of the meshes 1 to 4 for calculating the predicted value of the total demand at the specified date and time may be partially omitted as long as the reliability of the prediction described later is maintained.

<Decision of reliability>
When a plurality of predicted values related to future total demand are calculated, the reliability of the prediction is determined in step S4. Such determination of reliability is executed by the reliability determination unit 143 of the demand calculation unit 140.

  The reliability of the prediction is determined based on the distribution of the predicted value of the total demand at the specified date and time (August 1, 2014 at 0:00). In the present embodiment, the predicted value of total demand at the specified date and time is 27 to the fourth power. First, the reliability determination unit 143 obtains the arithmetic mean and standard deviation for the predicted value of total demand. In the present embodiment, it is assumed that the correspondence table indicating the correspondence relationship between the standard deviation value and the reliability is set in advance, and the reliability determination unit 143 refers to such a correspondence table and determines the reliability of the prediction. To decide. The correspondence between the standard deviation value and the reliability is, for example, as the standard deviation value increases from 0, the A rank (reliability: high), the B rank (reliability: medium), and the C rank (reliability). : Low).

  When the prediction reliability is determined for the specified date and time in this way, the reliability determination unit 143 determines the specified date and time as 1 August 2014, 1 o'clock, 2 o'clock on the same day ... As described above, the predetermined time interval (1 hour in the present embodiment) is changed toward the future, and the procedure for determining the reliability of prediction at the date and time is repeated. Then, by repeating such a determination procedure of the reliability, the reliability determination unit 143 determines the reliability for the prediction of the total demand in the predetermined area A up to several tens of hours ahead, thereby completing Step S4.

  Thus, the accuracy of the expected value of the total demand to be described later is improved by obtaining the reliability information for the prediction of the total demand.

  Note that the reliability of prediction may be determined based on the characteristics of the histogram (the height near the center and the extent of the skirt) as shown in FIGS.

<Determination of total demand forecast>
When the reliability is determined in step S4, the expected value of the total demand in the predetermined area A is determined in step S5. The determination of the expected value of the total demand is executed by the expected value determination unit 144 of the demand calculation unit 140.

  The expected value of the total demand is determined based on the reliability of the prediction at the designated date and time (August 1, 2014 at 0:00). For example, if the reliability at the designated date and time is rank A (reliability: high), the expected value determination unit 144 estimates the total demand by increasing the average value of the predicted value of the total demand at that date and time by 5%. Is rank B (reliability: medium), the total demand is expected to increase by 10% from the average value, and if the reliability is rank C (reliability: low), the total is increased by 30% from the average value. Expect demand. Alternatively, the prospective value determination unit 144 calculates the predicted values corresponding to the 55th percentile (median value), the 60th percentile, and the 70th percentile among all the predicted values according to the ranks A, B, and C, and estimates the total demand. It may be determined as a value. As another method, the expected value may be determined by a range (section), and the expected value determination unit 144 uses, for example, an average value (μ) and a standard deviation (σ), and is − (μ + 3σ) or more + (μ + 3σ). ) The following range may be determined as the estimated total demand. Furthermore, based on the concept of interval estimation in statistics, the expected value determination unit 144 may use, for example, a confidence interval for the confidence coefficient of 95% as the expected value of total demand.

  When the estimated value of the total demand at the specified date and time is determined in this way, the estimated value determining unit 144 sets the specified date and time at 1 o'clock on August 1, 2014, 2 o'clock on the same day, ... As shown in the diagram, the predetermined time interval (1 hour in the present embodiment) is changed toward the future, and the procedure for determining the expected value of the total demand at that date and time is repeated. Then, by repeating such a procedure for determining the expected value, the expected value determining unit 144 determines the expected value of the total demand in the predetermined area A up to several tens of hours ahead, and thereby step S5 is completed.

  Thus, by obtaining the estimated value of the total demand based on the reliability, the accuracy of the prediction is improved, and the power generation plan can be operated efficiently and stably.

  As described above, the demand calculation unit 140 uses the prediction formula indicating the relationship between the power demand at the power loads R1 to R4 and the weather information at the positions of the power loads R1 to R4, and the designated date and time at the positions of the power loads R1 to R4. And a plurality of predicted values related to the power demands of the power loads R1 to R4 at the designated date and time. The reliability determination part 143 determines the reliability with respect to the estimated electric power demand based on the distribution in a some predicted value. Therefore, the information regarding the reliability of prediction can be added to the prediction regarding power demand. This contributes to efficient and stable operation of the power generation plan in the meshes 1 to 4 and the predetermined area A.

  Further, the expected value determination unit 144 determines the expected value of the power demand based on the reliability, so that the prediction accuracy is improved and the power demand can be predicted in consideration of the weather risk.

  In addition, the prediction formula deriving unit 130 generates a prediction formula based on the weather observation information in the past predetermined period and the actual value of the power demand in the past predetermined period. The power demand can be predicted in consideration of the situation, and the prediction accuracy is improved.

  It is preferable that the power loads R1 to R4 are respectively provided in the meshes 1 to 4 (plural areas) in the predetermined area A, and the demand calculation unit 140 includes a first calculation unit 141 and a second calculation unit 142. The first calculation unit 141 calculates a plurality of area prediction values related to the power demand for each mesh 1 to 4, and the second calculation unit 142 based on the plurality of area prediction values calculated for each mesh 1 to 4, A plurality of regional prediction values related to the total power demand in the predetermined region A are calculated, and the reliability determination unit 143 determines the reliability of the predicted total demand based on the distribution of the plurality of regional prediction values. Thus, by calculating a plurality of regional prediction values, it becomes possible to predict the total demand in the predetermined region A in consideration of weather risk, and the reliability of the prediction is improved. This contributes to efficient and stable operation of the power generation plan in the predetermined area A.

  Each of the plurality of regional prediction values is preferably a sum obtained by adding any one of the plurality of regional prediction values calculated for each of the meshes 1 to 4 over all the meshes. By calculating such a sum, it is expected that the errors included in the predicted values for each mesh 1 to 4 are canceled out, so that it is possible to predict the total demand in consideration of the weather risk and the reliability of the prediction The degree is improved.

  Further, since the distribution of the predicted values is indicated by the arithmetic mean and the standard deviation, the variation of the predicted values is expressed by numerical values, and an objective and highly reliable demand prediction is possible.

  In addition, since the plurality of weather forecast values are values based on the ensemble forecast, the degree of variation in the forecast values is reflected in the forecast value of the power demand, so that it is possible to predict the power demand in consideration of reliability.

  In addition, since multiple weather forecast values include multiple forecast values related to at least one of temperature, humidity, wind speed, precipitation, solar radiation, and cloud cover, the power demand that is closely related to these weather conditions can be accurately As well as being able to predict well, the reliability of the prediction can also be obtained.

  In addition, the said embodiment is for making an understanding of this invention easy, and is not for limiting and interpreting this invention. The present invention can be changed and improved without departing from the gist thereof, and the present invention includes equivalents thereof.

DESCRIPTION OF SYMBOLS 100 Prediction apparatus 110 Transmission / reception part 120 Storage part 130 Prediction formula derivation part 140 Demand calculation part R1-R4 Electric power load M1-M4 Measurement apparatus W1-W4 Weather observation apparatus WF Weather forecast apparatus

Claims (10)

A demand prediction device for predicting power demand,
Based on a prediction formula indicating a relationship between power demand in the power load and weather information at the position of the power load, and a plurality of weather forecast values for the specified date and time at the position of the power load, the specified date and time A demand calculation unit for calculating a plurality of predicted values related to the power demand of the power load at
A reliability determination unit that determines the reliability of the predicted power demand based on the distribution of the plurality of predicted values;
A demand prediction apparatus comprising:   The demand forecasting device according to claim 1, further comprising a forecast value determining unit that determines a forecast value of power demand based on the reliability.   3. A prediction formula deriving unit that generates the prediction formula based on weather observation information in a past predetermined period and an actual value of power demand in the past predetermined period is further included. The demand forecasting device described in 1. The power load is provided in each of a plurality of areas in a predetermined area,
The demand calculation unit
A first calculation unit that calculates a plurality of predicted area values related to power demand for each of the plurality of areas;
A second calculation unit that calculates a plurality of regional prediction values related to total power demand in the predetermined region based on the plurality of regional prediction values calculated for each of the plurality of regions;
With
The said reliability determination part determines the reliability with respect to the estimated total demand based on distribution of these several regional prediction values. The demand prediction apparatus in any one of Claims 1-3 characterized by the above-mentioned.   Each of the plurality of regional prediction values is a sum obtained by adding any one of the plurality of region prediction values calculated for each of the plurality of regions over all the plurality of regions. 4. The demand prediction apparatus according to 4.   The demand prediction apparatus according to claim 1, wherein the distribution is indicated by an arithmetic average and a standard deviation.   The demand forecasting device according to claim 1, wherein the plurality of weather forecast values are values based on an ensemble forecast. The plurality of weather forecast values include a plurality of forecast values related to at least one of temperature, humidity, wind speed, precipitation, solar radiation, and cloud cover,
The demand prediction device according to any one of claims 1 to 7, wherein A demand prediction method for predicting power demand,
Based on a prediction formula indicating a relationship between power demand in the power load and weather information at the position of the power load, and a plurality of weather forecast values for the specified date and time at the position of the power load, the specified date and time Calculating a plurality of predicted values related to the power demand of the power load at
Determining a confidence level for the predicted power demand based on the distribution in the plurality of predicted values;
Demand forecasting method characterized by this. In a demand forecasting device that forecasts power demand,
Based on a prediction formula indicating a relationship between power demand in the power load and weather information at the position of the power load, and a plurality of weather forecast values for the specified date and time at the position of the power load, the specified date and time A function of calculating a plurality of predicted values related to the power demand of the power load at
A function of determining reliability for the predicted power demand based on the distribution in the plurality of predicted values;
A program that executes

Images