ENERGY PREDICTION DEVICE, ENERGY PREDICTION METHOD, AND ENERGY PREDICTION PROGRAM

Abstract

To provide a technique capable of improving prediction accuracy of a prediction value of energy.SOLUTION: An energy prediction device acquires a plurality of temporary prediction values of energy predicted by a plurality of prediction methods different from each other, changes a ratio of at least one temporary prediction value among the plurality of temporary prediction values according to prediction time, and calculates a prediction value of energy by blending the plurality of temporary prediction values.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to an energy prediction device, an energy prediction method, and an energy prediction program. [Background technology]

[0002] Conventionally, prediction of renewable energy or prediction of energy demand have been considered. For example, Patent Document 1 describes a system that creates a power generation conversion model based on past actual values ​​of meteorological elements and actual values ​​of renewable energy power generation, and calculates a predicted value of future renewable energy power generation by inputting a predicted value of future meteorological elements into the power generation conversion model. Patent Document 2 describes a system that generates a demand forecast model that represents the relationship between fluctuation factor data and power demand data for each of a plurality of customer groups based on teacher data including actual data of fluctuation factors of power demand and actual data of power demand. Patent Document 3 describes an apparatus that constructs a forecast model that predicts the coefficient of angular frequency by inputting the Fourier transform result of actual data of power demand and the actual data, and calculates power demand forecast data by inputting forecast data into the forecast model and performing an inverse Fourier transform on the calculated coefficient forecast data.

[0003] Patent Literature 4 describes an apparatus that blends a plurality of power demand forecast values ​​for a target forecast date, which are predicted by each of a plurality of forecast models, by taking a weighted average at a predetermined ratio to obtain a power demand forecast value for the target forecast date. Patent Literature 5 describes a system that obtains a past predicted power amount for each of a plurality of forecasting means for calculating the amount of power, and selects a forecasting means that reduces error based on the past predicted power amount and the past actual power amount. Patent Literature 6 describes an apparatus that obtains a forecast value for each method using a plurality of forecasting methods, obtains a weight for each forecasting method based on the error from the actual power value for a specific day, and calculates a power demand forecast value for a consumer based on the forecast value for each method and the weight. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2022-121028 A [Patent Document 2] Patent Publication No. 2022-84328 [Patent Document 3] Patent No. 6915156 [Patent Document 4] Patent No. 7062144 [Patent Document 5] Patent No. 5734325 [Patent Document 6] Patent No. 7075772 [Patent Document 7] Patent No. 6705716 Summary of the Invention [Problem to be solved by the invention]

[0005] There is a need to further improve the accuracy of energy forecast values.

[0006] An object of the present disclosure is to provide a technique capable of improving the prediction accuracy of a predicted value of energy. [Means for solving the problem]

[0007] An energy prediction device according to one aspect of the present disclosure obtains a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, changes a ratio of at least one of the plurality of tentative prediction values ​​according to the time of prediction, and blends the plurality of tentative prediction values ​​to calculate a predicted value of energy.

[0008] An energy prediction method according to one aspect of the present disclosure includes the steps of obtaining a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, and calculating a predicted value of energy by blending the plurality of tentative prediction values ​​by varying a ratio of at least one of the plurality of tentative prediction values ​​according to a time to be predicted.

[0009] An energy prediction program according to one aspect of the present disclosure causes a computer to execute the steps of obtaining a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, and calculating a predicted value of energy by changing a ratio of at least one of the plurality of tentative prediction values ​​according to a prediction time and blending the plurality of tentative prediction values.

[0010] In the energy prediction device, the energy prediction method, and the energy prediction program, a predicted value of energy is calculated based on a plurality of provisional predicted values ​​of energy predicted by a plurality of different prediction methods. At this time, the distribution ratio of the plurality of provisional predicted values ​​is changed and blended according to the prediction time. Here, the prediction accuracy of the predicted value of energy may change according to the prediction time. By combining the provisional predicted values ​​by the plurality of different prediction methods according to the prediction time, it is possible to improve the prediction accuracy of the predicted value of energy compared to a single prediction method.

[0011] At least one of the plurality of tentative predicted values ​​may be calculated by a prediction method using an actual energy value. The tentative predicted value based on the actual energy value is reflected in the predicted value, thereby improving prediction accuracy of the predicted energy value.

[0012] A plurality of provisional prediction values ​​may be blended so that the ratio of provisional prediction values ​​calculated by a prediction method using actual energy values ​​is dominant from the current time to a specified future time. The nearest future time can be said to be a prediction period in which the prediction accuracy of provisional prediction values ​​based on actual energy values ​​is high or is expected to be high. By making the ratio of provisional prediction values ​​based on actual energy values ​​dominant from the current time to a specified future time, the prediction accuracy of the predicted energy values ​​is further improved.

[0013] The energy may be power generated by photovoltaic power generation. At least one of the provisional predicted values ​​may be calculated by a prediction method using weather forecast data. The provisional predicted value based on the weather forecast data is reflected in the predicted value of the power generated by photovoltaic power generation, thereby improving the prediction accuracy of the predicted value of the energy.

[0014] A plurality of provisional forecast values ​​may be blended so that the ratio of provisional forecast values ​​calculated by a forecasting method using weather forecast data becomes dominant from a predetermined future time onward. Future times far from the current time can be considered to be forecast periods in which the prediction accuracy of provisional forecast values ​​based on weather forecast data is high or is expected to be high. By making the ratio of provisional forecast values ​​based on weather forecast data dominant from a future time far from the current time onward, the prediction accuracy of the energy forecast value is further improved.

[0015] A prediction error band of the predicted energy value may be calculated by further acquiring a corresponding tentative prediction value upper limit and a tentative prediction value lower limit for each of the plurality of tentative prediction values ​​and blending the corresponding tentative prediction value upper limit and the tentative prediction value lower limit in accordance with a ratio of the plurality of tentative prediction values. By calculating the prediction error band, the convenience of a user who uses the predicted energy value is improved.

[0016] The prediction methods may use different algorithms for prediction. By using different algorithms for each of the prediction methods, it is possible to improve the prediction accuracy of the predicted energy value.

[0017] The prediction methods may use different data resources for prediction. The prediction accuracy of the predicted value may reflect the time characteristics of the different data resources. This makes it possible to improve the prediction accuracy of the energy prediction value.

[0018] The prediction methods may have different prediction periods. By using different prediction methods for each prediction period, it is possible to improve the prediction accuracy of the predicted energy value. In one example, a tentative predicted value based on a prediction period from the current time to a first future time is combined with a tentative predicted value based on a prediction period from the current time to a second future time after the first future time. In this case, it is possible to improve the prediction accuracy of the predicted energy value from the first future time to the second future time.

[0019] The multiple prediction methods may differ from each other in the number of predicted values ​​in the prediction period. By using multiple prediction methods that differ according to the number of predicted values ​​(prediction scores), it is possible to improve the prediction accuracy of the energy predicted value. In one example, a tentative prediction value based on prediction scores for a 30-minute interval and a tentative prediction value based on prediction scores for a 1-hour interval are combined. Since the characteristics due to the time interval are reflected in the tentative prediction value, it is possible to improve the prediction accuracy of the energy predicted value. Effect of the Invention

[0020] According to the present disclosure, it is possible to provide a technique capable of improving the prediction accuracy of a predicted value of energy. [Brief description of the drawings]

[0021] [Figure 1] FIG. 1 is a schematic diagram of a power supply system. [Diagram 2] FIG. 2 is a block diagram illustrating the functional configuration of the EMS according to the first embodiment. [Diagram 3] FIG. 3 is a conceptual diagram illustrating prediction according to the persistence model. [Figure 4] FIG. 4 is a conceptual diagram illustrating prediction by an AR model. [Diagram 5] FIG. 5 is a conceptual diagram illustrating a prediction based on a clear day power generation model. [Figure 6] FIG. 6 is a diagram illustrating the concept of blending tentative predictions. [Figure 7]FIG. 7 is a flow chart illustrating an example of an energy prediction method. [Figure 8] FIG. 8 is a block diagram illustrating the functional configuration of the EMS according to the second embodiment. [Figure 9] Fig. 9(a) is a conceptual diagram showing the blending ratio function α, Fig. 9(b) is a conceptual diagram showing the blending ratio function β, and Fig. 9(c) is a conceptual diagram showing the blending ratio function γ. [Figure 10] FIG. 10 is a flow chart illustrating another example of a method for energy prediction. [Figure 11] FIG. 11 is a diagram illustrating an example of a hardware configuration related to a power supply system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same elements are given the same reference numerals, and duplicated explanations will be omitted.

[0023] [Power supply system] The power supply system according to the present disclosure calculates a predicted value of energy based on a plurality of tentative predicted values ​​of energy predicted by a plurality of prediction methods. For example, the power supply system is applied to predict the power generation of a photovoltaic power generation facility in a microgrid. In the present disclosure, the unit of the predicted value may be power [kW] or power amount [kWh]. The predicted value of the generated power is used, for example, to optimize the charging and discharging of a storage battery system. The optimization of charging and discharging means planning the amount of charging and discharging power of the storage battery system so that the amount of power between the microgrid and the power system is economically optimized. In addition, the target to be optimized is not limited to the charging and discharging power of the storage battery. For example, the predicted value may be used to determine the power consumption of a water electrolysis hydrogen production device so that the hydrogen production cost is economically optimized. In addition, when the prediction target is power demand, the predicted value may be used to determine the power generation or start / stop time of a generator such as a diesel engine so that it is economically optimal. Furthermore, the predicted values ​​of photovoltaic power generation and power demand may be used to plan the operation of a group of devices including a storage battery, a hydrogen production device, and a generator. The use of the predicted value is not limited to formulating the operation plan of these devices, and may be used for bidding on the electricity market or applying for self-consignment of electricity, command decision of demand response to avoid tight electricity supply, etc. The predicted value may be used in the form of a distribution service of the predicted value that provides the predicted value to a third party such as a microgrid operator. In this case, the distributor of the predicted value is not particularly aware of the use of the predicted value.

[0024] FIG 1 is a schematic configuration diagram of a power supply system 1 according to one embodiment. The power supply system 1 includes a microgrid 2 and an EMS (energy management system) 3. The microgrid 2 and the EMS 3 are connected to each other so as to be able to communicate with each other. The EMS 3 may be a component of the microgrid 2.

[0025] The microgrid 2 is connected to an external power system 90. The microgrid 2 can transmit and receive power to and from the power system 90. The microgrid 2 includes a solar power generation facility 21, a generated power measurement unit 22, a storage battery system 23, a connection unit 24, a received power measurement unit 25, and a transmitted power measurement unit 26.

[0026] The solar power generation equipment 21 is an example of a renewable energy power generation device. The solar power generation equipment 21 is equipment that generates power using sunlight (Photovoltaic: PV). The solar power generation equipment 21 includes a solar panel 21a and a power conditioner (Power Conditioning System: PCS) not shown. The PCS converts direct current into alternating current. The solar power generation equipment 21 is connected to a connection unit 24. The solar power generation equipment 21 outputs generated power to the connection unit 24.

[0027] The generated power measurement unit 22 measures the actual amount of power generated in the solar power generation facility 21. The method for measuring the actual amount of power generated is not limited. For example, the generated power measurement unit 22 may measure the power output from the solar power generation facility 21 to the connection unit 24.

[0028] The battery system 23 is an example of an energy storage device. The battery system 23 includes a battery. The battery is a secondary battery such as a lithium ion battery, a lead acid battery, or a redox flow battery. The battery may be an energy storage device such as a flywheel, a compressed air energy storage (CAES) facility, or a large-capacity capacitor. The battery system 23 includes a battery PCS that converts the direct current of the battery to alternating current, and a device for monitoring the remaining battery charge. The battery system 23 can function as a power adjustment device when adjusting the exchange of power between the microgrid 2 and the outside.

[0029] The connection unit 24 is, for example, a distribution board. The connection unit 24 controls the power distribution to the storage battery system 23 and the power grid 90. For example, the connection unit 24 controls the power distribution based on an instruction from the EMS 3.

[0030] The received power measuring unit 25 measures the received power from the external power system 90. The transmitted power measuring unit 26 measures the transmitted power to the external power system 90.

[0031] [First embodiment] 2 is a block diagram illustrating a functional configuration of the EMS 3 according to the first embodiment. The EMS 3 functions as an energy prediction device in the present disclosure. The EMS 3 is communicatively connected to a weather forecast service 95 via a network N such as the Internet. The EMS 3 is communicatively connected to a generated power measurement unit 22 of the microgrid 2.

[0032] The weather forecast service 95 periodically calculates weather data based on meteorological satellite data or observation data from various locations, using weather simulations by supercomputers or numerical correction techniques using artificial intelligence. Examples of weather data include, but are not limited to, temperature, humidity, wind speed, wind direction, precipitation, snowfall, snow depth, total solar radiation, and cloud cover. In one example, the weather forecast service 95 calculates weather data for the next 1 to 3 days in an area including the latitude and longitude of the microgrid 2.

[0033] The EMS 3 includes an external communication unit 31, a first prediction unit 32, a performance database 33, a second prediction unit 34, and a calculation unit 35. Fig. 2 illustrates only the functions related to the energy prediction according to the present disclosure among the various functions of the EMS 3. That is, Fig. 2 omits illustration of functions such as a user interface, a report function, a communication function with the storage battery system 23, and a function for creating a plan for the amount of charge and discharge power of the storage battery.

[0034] The external communication unit 31 acquires, for example, weather forecast data from a weather forecast service 95. The external communication unit 31 periodically acquires the weather forecast data by a communication protocol such as FTP (File Transfer Protocol) or HTTP (Hypertext Transfer Protocol). The timing of acquisition may be, for example, once a day, four times a day, or 48 times a day, but is not limited thereto.

[0035] The first prediction unit 32 calculates a tentative predicted value of energy by a first prediction method. For example, the first prediction method may be an energy prediction method using weather forecast data. The first prediction unit 32 may output a tentative predicted value upper limit and a tentative predicted value lower limit in addition to the tentative predicted value. The tentative predicted value upper limit and tentative predicted value lower limit are, for example, ranges within which the value of the prediction target falls with a predetermined accuracy (e.g., 80%).

[0036] The performance database 33 is a non-transitory storage medium or storage device that stores the performance values ​​of the generated power. The performance database 33 stores the performance values ​​of the generated power measured by the generated power measurement unit 22 in association with the time. In other words, the performance database 33 stores the performance values ​​of the generated power from the present time onwards over time. The performance database 33 may be constructed as a single database or may be a collection of multiple databases. The location where the performance database 33 is installed is not limited. For example, the performance database 33 may be provided in a computer system separate from the power supply system 1.

[0037] The second prediction unit 34 calculates a provisional prediction value of energy by a second prediction method. For example, the second prediction method may be an energy prediction method using an actual value of power generation. The second prediction unit 34 may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value.

[0038] The calculation unit 35 obtains a plurality of tentative predicted values ​​of energy predicted by a plurality of prediction methods different from each other. The plurality of prediction methods may be different from each other in any one or more of the following: a prediction algorithm is different, a data resource is different, a prediction period is different, or a prediction score, which is the number of predicted values ​​within a prediction period, is different. The prediction period is a time range for which a predicted value is obtained, and is also called a prediction interval.

[0039] Among the multiple prediction methods, for example, even if the algorithm is the same, when different data resources are used, they are multiple prediction methods different from each other. For example, the external communication unit 31 of the EMS 3 may acquire weather forecast data from the weather forecast service 95 and a weather forecast service different from the weather forecast service 95. The first prediction unit 32 may calculate multiple provisional predicted values ​​W1 and W2 based on each weather forecast data. In another example, the first prediction method may be a prediction method whose prediction period is farther from the current time than the second prediction method. In other words, the second prediction method may be a prediction method whose prediction period is closer to the current time than the first prediction method. In yet another example, the first prediction method may be a prediction method whose prediction score is smaller than the second prediction method. In other words, the second prediction method may be a prediction method whose prediction score is larger than the first prediction method. The prediction score can be rephrased as the resolution of the prediction period. The prediction score may be expressed, for example, as a 30-minute interval or an hourly interval.

[0040] The calculation unit 35 changes the ratio of at least one of the multiple provisional predicted values ​​according to the time to be predicted, and blends the multiple provisional predicted values ​​to calculate the predicted value of energy. For example, the calculation unit 35 calculates the predicted value B of energy based on the provisional predicted value W of energy predicted by the first prediction unit 32 and the provisional predicted value T of energy predicted by the second prediction unit 34. At this time, the calculation unit 35 changes the allocation ratio of the provisional predicted values ​​W and T according to the time to be predicted. In other words, the calculation unit 35 calculates the predicted value B of energy using a ratio calculation function in which the time to be predicted is an independent variable and the allocation ratio is a dependent variable.

[0041] The calculation unit 35 may blend the prediction errors. For example, the calculation unit 35 may further acquire a corresponding tentative prediction value upper limit and a tentative prediction value lower limit for each of the multiple tentative prediction values. The calculation unit 35 may calculate a prediction error range of the energy prediction value B by blending the corresponding tentative prediction value upper limit and tentative prediction value lower limit according to the ratio of the multiple tentative prediction values.

[0042] [First prediction method] The first prediction unit 32 calculates the provisional prediction value W by a prediction method using weather forecast data as the first prediction method. The first prediction unit 32 extracts weather data for a prediction period from the weather forecast data. The weather data includes, but is not limited to, the amount of solar radiation and the temperature. The first prediction unit 32 calculates the provisional prediction value W by the following formula (1) using the amount of solar radiation and the temperature.

[0043]

number

[0044] In formula (1), y W is the provisional predicted value W [kW]. i W is the predicted time (i=1,2,...,N W ). N W represents the predicted score of the tentative predicted value W. total is the overall correction coefficient [-]. The overall correction coefficient is a parameter that represents all efficiency reductions other than those due to temperature rise. Factors that cause efficiency reduction include cell deterioration, dirt on the array surface, circuit loss, and PCS conversion loss.

[0045] K total is usually between 0.6 and 0.9. PT is the temperature correction coefficient [-]. The temperature correction coefficient represents the decrease in efficiency due to an increase in temperature. K PT is the cell temperature T cell It depends on [℃] and is calculated by the following formula (2).

[0046]

number

[0047] In formula (2), α Pmax is the maximum output temperature coefficient [% / ℃]. α Pmax The cell temperature T varies depending on the type of cell crystal, and for example, in some crystal systems it takes a negative value of about -0.4 to -0.5. cell[℃] may be calculated separately from weather forecast data (temperature, solar radiation, wind speed, etc.), or may be the same as the temperature. cell [℃] is explained as being the same as air temperature.

[0048] In formula (1), P AS is the PV array output (total capacity of the PV array) [kW] under the reference condition. I is the total solar radiation [W / m 2 The global solar radiation may be separated into a direct component and a scattered component by a method such as the Erbs method, and converted into an oblique solar radiation of the PV array (oblique solar radiation).

[0049] [Second prediction method] The second prediction unit 34 calculates the tentative prediction value T by a prediction method using the actual value of the power generation as the second prediction method. The second prediction unit 34 acquires actual values ​​for a certain period of time in the past from the actual database 33. An example of the certain period of time in the past is, but is not limited to, a three-hour period as an average value at 30-minute intervals. The actual values ​​for the certain period of time in the past can be considered as time-series data. The second prediction unit 34 calculates the tentative prediction value T by a prediction model using the time-series data. Examples of the prediction model include, but are not limited to, a sustained model, an autoregressive model (AR model), and a sunny day power generation model.

[0050] FIG. 3 is a conceptual diagram illustrating prediction by the persistence model. FIG. 3 shows actual values ​​P3a and predicted values ​​P3b. The persistence model is a model that predicts by assuming that future values ​​are the same as the current value. In the persistence model, all actual values ​​P3a other than the final value are ignored. For example, the second prediction unit 34 sets the tentative predicted value T as the final value (latest value) of the actual values ​​P3a over the entire prediction period from the current time onwards. The persistence model is used as the simplest method, and is particularly effective when the desired prediction period is short. The persistence model is also used as a standard prediction method when evaluating the accuracy of other prediction methods.

[0051] FIG. 4 is a conceptual diagram illustrating prediction by an AR model. FIG. 4 shows actual values ​​P4a and predicted values ​​P4b. The AR model is expected to have high prediction accuracy for values ​​that have a highly autoregressive (periodic or oscillatory) trend, such as wind power generation or power demand. In one example, power demand has a high tendency to have a 24-hour periodicity. Time series data {y i}(i=1, ,N), the AR model is i is expressed by the following equation (3).

[0052]

number

[0053] In equation (3), m is the order of the autoregression. i is the coefficient. v n has mean 0 and variance σ 2 The time series data of the actual values ​​{y i}(i=1, ,N), we choose the order m and coefficient a of the AR model so that the Akaike Information Criteria (AIC) is minimized. i (i=1, ,m) is determined. AIC is a type of statistic. The smaller the AIC, the better the balance between the model's complexity and its fit to the data. The AIC of an AR model of order m is expressed by the following formula (4).

[0054]

number

[0055] The second prediction unit 34 predicts the PV power generation amount in the prediction period using the obtained AR model. n is usually treated as zero. For example, the predicted value y^ n+1 is the time series data of the actual values ​​{y i}(i=1, ,N) is used to calculate the following equation (5).

[0056]

number

[0057] Furthermore, the predicted value y^ for the next time n+2 is calculated by the following formula (6).

[0058]

number

[0059] The second prediction unit 34 predicts future predicted values ​​by repeatedly calculating formula (6). The final predicted value of the AR model may be limited to within a range of maximum and minimum values ​​specified by the user of the EMS 3. This makes it possible to prevent the predicted value of the AR model from being too large or too small.

[0060] FIG. 5 is a conceptual diagram illustrating a prediction using the clear day power generation model. FIG. 5 shows an actual value P5a and a predicted value P5b. The clear day power generation model is applied only to the prediction of PV power generation. The clear day power generation model is not applied to energy demand prediction and wind power generation prediction, etc. The clear day power generation model can capture the trend of PV power generation having characteristics such as sunrise and sunset. The clear day power generation model is a model that uses a value converted into PV power generation power based on theoretically calculated extraterrestrial solar radiation. The clear day power generation model calculates a future predicted value P5b assuming that the ratio between the theoretically calculated PV power generation power and the actual value P5a is constant. The second prediction unit 34 calculates the predicted value of PV power generation power using the following formula (7).

[0061]

number

[0062] In equation (7), k is the current time. pv real (k) is the actual power generation at time k. pv sunny(t) is the PV power generation on a clear day at time t. p^ pv (t|k) is the predicted PV power generation value at future time t at the current time k.

[0063] In formula (7), p pv real (k) / p pv sunny (k) can be considered as an index similar to the clear weather index in calculating solar radiation. pv sunny (t) may be PV power generation power calculated using a method such as the Erbs method, with the amount of solar radiation on a horizontal surface outside the atmosphere being taken as the amount of solar radiation on a horizontal surface globally.

[0064] At night, etc., pv sunny Since (k) becomes zero, p pv real (k) / p pv sunny (k) becomes indefinite. pv sunny If (k) is small enough, p pv real (k) / p pv sunny (k) may be set to 0.5, or may not be used in the blending by the calculation unit 35.

[0065] p^ pv If (t|k) is less than or equal to zero, p^ pv (t|k) may be set to zero. pv When (t|k) is equal to or greater than the rated value of the PV power generation, p^ pv (t|k) may be limited to a nominal value.

[0066] [blend] Fig. 6 is a diagram illustrating the concept of blending tentative predicted values. Fig. 6 shows an actual energy value K in the past from the current time, and a predicted energy value in the future from the current time. The actual energy value K is, for example, the generated power measured by the generated power measurement unit 22. In Fig. 6, a tentative predicted value W, a tentative predicted value T, and a predicted value B are shown as predicted values ​​of future energy.

[0067] The calculation unit 35 changes the allocation ratio of the provisional predicted value W and the provisional predicted value T according to the time of prediction, and calculates the predicted value B. For example, the calculation unit 35 prioritizes the provisional predicted value T as a predicted value for the future close to the current time within the prediction period of the predicted value B. The calculation unit 35 prioritizes the provisional predicted value W as a predicted value for the future far from the current time within the prediction period of the predicted value B.

[0068] For example, the calculation unit 35 increases the allocation ratio of the provisional predicted value T and decreases the allocation ratio of the provisional predicted value W at time t1 in the near future from the current time to calculate the predicted value B. The calculation unit 35 decreases the allocation ratio of the provisional predicted value T and increases the allocation ratio of the provisional predicted value W at time t2 in the far future from the current time to calculate the predicted value B. In other words, the calculation unit 35 calculates a predicted value B that is closer to the provisional predicted value T than the provisional predicted value W as the time gets closer to the current time. The calculation unit 35 calculates a predicted value B that is closer to the provisional predicted value W than the provisional predicted value T as the time gets farther from the current time.

[0069] Let the number of data points for the predicted value X (X=W,T,B) be N. x The predicted time is as follows:

[0070]

number

[0071] The predicted value is y x The tentative predicted values ​​W and T are given, and the predicted value B is k B} k=1 NAssume that only the current time is given. The interval of the predicted time of the predicted value B does not have to be constant. In the future close to the current time, the predicted time interval may be set to be short. In the future far from the current time, the predicted time interval may be set to be long. The predicted time interval may be set to be, for example, 5 minute intervals up to 1 hour ahead, 30 minute intervals from 1 hour to the 1st, and 3 hour intervals from the 1st to the 3rd. The predicted times may all be different for the provisional predicted value W, the provisional predicted value T, and the predicted value B.

[0072] The calculation unit 35 calculates the value y of the predicted value B in the following procedure. B (t k B ) is calculated. If the predicted time of the tentative predicted value W is {t k B} k=1 N In the case of extrapolation, the first or last value of the tentative prediction value W is used. That is, t i B <t 1 W Then the following equation (8) is used:

[0073]

number

[0074]

number

[0075] Then the following equation (9) is used:

[0076]

number

[0077] The calculation unit 35 also performs a similar resampling process on the provisional predicted value T. The calculation unit 35 calculates a predicted value B by the following equation (10) using the provisional predicted value W and the provisional predicted value T that have been subjected to the resampling process.

[0078]

number

[0079] In formula (10), α is a blending ratio function that determines the distribution ratio of the tentative forecast value. α is calculated by the following formula (11).

[0080]

number

[0081] In equation (11), λ is the time constant specified by the user. end is the last time of a given actual value. When the first sampling time of the predicted value B coincides with the last sampling time of the actual value, the predicted value B can be expressed as the following equation (12).

[0082]

number

[0083] The predicted value B approaches the tentative predicted value W as the prediction time moves into the future. The speed of the approach depends on the time constant λ. For example, at time t end At time t +λ, the prediction B is a blend of 63.8% of the tentative prediction W and 36.2% of the tentative prediction T. end At time t +2λ, the prediction B is a blend of 86.5% of the tentative prediction W and 13.5% of the tentative prediction T. end At time +5λ, the prediction B is blended with less than 1% of the tentative prediction T.

[0084] The operation of the EMS 3 will be described, along with an example of an energy prediction method, with reference to Fig. 7. Fig. 7 is a flow chart showing an example of the operation of the EMS 3 as a process flow S1.

[0085] In step S11, the EMS 3 acquires first data. For example, the external communication unit 31 acquires weather forecast data from the weather forecast service 95 as the first data.

[0086] In step S12, the EMS 3 calculates a first tentative predicted value of energy by a first prediction method. For example, the first prediction unit 32 calculates a tentative predicted value W as the first tentative predicted value by a first prediction method using weather forecast data. For example, the first prediction unit 32 calculates the tentative predicted value W by using equations (1) and (2) or the like.

[0087] In step S13, the EMS 3 acquires second data. For example, the second prediction unit 34 acquires a record value of the generated power from the record database 33 as the second data.

[0088] In step S14, the EMS 3 calculates a second provisional predicted value of energy by a second prediction method. For example, the second prediction unit 34 calculates a provisional predicted value T as the second provisional predicted value by a second prediction method using an actual value of the power generation. For example, the second prediction unit 34 calculates the provisional predicted value T by using a prediction model such as a sustained model, an AR model, and a clear day power generation model. For the AR model, for example, formulas (3) to (6) are used. For the clear day power generation model, for example, formula (7) is used.

[0089] In step S15, the EMS 3 changes the ratio of at least one of the multiple provisional predicted values ​​according to the time of prediction, and blends the multiple provisional predicted values ​​to calculate a predicted value of energy. For example, the calculation unit 35 obtains multiple provisional predicted values ​​(first provisional predicted value and second provisional predicted value) of energy predicted by multiple prediction methods different from each other. For example, the calculation unit 35 changes the ratio of the first provisional predicted value and the second provisional predicted value according to the time of prediction, and blends them to calculate a predicted value B of energy. For example, the calculation unit 35 calculates the predicted value B using equations (8) to (12) or the like.

[0090] In the process flow S1, steps S11 and S12 and steps S13 and S14 may be executed in parallel or in series. In relation to steps S12, S14, and S15, the EMS3 may blend the prediction errors. For example, in step S12, the first prediction unit 32 may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value W. In step S14, the second prediction unit 34 may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value T. In step S15, the calculation unit 35 may further obtain a corresponding provisional prediction value upper limit and a provisional prediction value lower limit for each of the multiple provisional prediction values ​​W and T. The calculation unit 35 may calculate a prediction error range of the energy prediction value B by blending the corresponding provisional prediction value upper limit and provisional prediction value lower limit according to the ratio of the multiple provisional prediction values ​​W and T. For example, the calculation unit 35 may calculate the prediction error range using the formula (10). In formula (10), the calculation unit 35 may apply a blend ratio function α to the provisional predicted value upper limit and the provisional predicted value lower limit corresponding to the provisional predicted value T. The calculation unit 35 may apply (1-α) to the provisional predicted value upper limit and the provisional predicted value lower limit corresponding to the provisional predicted value W.

[0091] [Second embodiment] FIG. 8 is a block diagram illustrating a functional configuration of an EMS3A according to the second embodiment. The EMS3A functions as an energy prediction device like the EMS3. Below, differences between the EMS3A and the EMS3 will be mainly described. The EMS3A differs from the EMS3 in that it is connected to an aerial photography device 96 that photographs the sky (all sky) and in that it includes a third prediction unit 36. While the EMS3 blends two provisional prediction values, the EMS3A blends three provisional prediction values. The number of provisional prediction values ​​to be blended may be three or more. The aerial photography device 96 may be, for example, a fisheye camera or a device that acquires satellite images.

[0092] The third prediction unit 36 ​​calculates a tentative predicted value of energy by a third prediction method. For example, the third prediction method may be an energy prediction method using video data of the sky. The third prediction unit 36 ​​may output a tentative predicted value upper limit and a tentative predicted value lower limit in addition to the tentative predicted value.

[0093] The third prediction unit 36 ​​calculates the tentative predicted value C by a prediction method using video data of the sky as the third prediction method. For example, the third prediction unit 36 ​​acquires video data of the sky from the aerial photography device 96. The third prediction unit 36 ​​calculates the tentative predicted value C of the future solar radiation or PV power generation power from the video data of the sky by a prediction method using machine learning or deep learning based on the fluctuation of cloud cover, etc.

[0094] The calculation unit 35 calculates the predicted value B based on the provisional predicted value W, the provisional predicted value T, and the provisional predicted value C. For example, the calculation unit 35 calculates the predicted value B by the following equation (13).

[0095]

number

[0096] In equation (13), α, β, and γ are blending ratio functions, respectively. In equation (13), it is always assumed that α≧0, β≧0, γ≧0, and α+β+γ>0. FIG. 9(a) is a conceptual diagram showing blending ratio function α. ​​FIG. 9(b) is a conceptual diagram showing blending ratio function β. FIG. 9(c) is a conceptual diagram showing blending ratio function γ. In FIG. 9(a), FIG. 9(b), and FIG. 9(c), the horizontal axis represents time, and the vertical axis represents each value α(t i B ), β(t i B ) and γ(t i B 9(a), the blending ratio function α is a monotonically decreasing function. As shown in FIG. 9(b), the blending ratio function β is a convex function. As shown in FIG. 9(c), the blending ratio function γ is a monotonically increasing function. In FIG. 9(a), FIG. 9(b), and FIG. 9(c), tend indicates the current time, and t end +1hour indicates 1 hour later, t end +5hour indicates 5 hours later. end In the blending ratio function α, the value α(t i B ) becomes dominant, and the blending ratio function β and γ value β(t i B ) and γ(t i B ) is zero or nearly zero. t end At +1 hour, the blend ratio function α and β values ​​α(t i B ) and β(t i B ) becomes dominant, and the value of the blending ratio function γ γ(t i B ) is zero. t end At +5 hours, the value of the blend ratio function α(t i B ) is almost zero, and the blending ratio functions β and γ values ​​β(t i B ) and γ(t i B ) becomes dominant.

[0097] In formula (13), the allocation ratios of the provisional predicted values ​​T, C, and W are changed depending on how far the predicted time is from the current time. For example, the calculation unit 35 blends the provisional predicted values ​​T, C, and W according to the respective allocation ratios corresponding to one hour later to calculate the predicted value B. In another example, the calculation unit 35 blends the provisional predicted values ​​T, C, and W according to the respective allocation ratios corresponding to five hours later to calculate the predicted value B. The respective allocation ratios are determined so that the allocation ratios of the provisional predicted values ​​T, C, and W are high for a prediction period with high prediction accuracy or a prediction period expected to be high for each of the provisional predicted values ​​T, C, and W. In other words, the allocation ratios of the corresponding provisional predicted values ​​are determined so that the allocation ratios are high for a prediction period in which each prediction method is good at it.

[0098] In the power prediction of PV power generation, the tentative predicted value T, the tentative predicted value C, and the tentative predicted value W may have different prediction accuracy depending on the time of prediction. For example, the tentative predicted value T based on the actual value of the power generation tends to have good prediction accuracy for about one hour from the current time. The tentative predicted value C based on cloud cover tends to have good prediction accuracy for about one to five hours from the current time. The tentative predicted value W based on weather data tends to have good prediction accuracy for about six to 48 hours (two days). In other words, the nearest future time of about one hour from the current time is a prediction period in which the prediction accuracy of the tentative predicted value T is high, or is expected to be high. The future time of about one hour to five hours is a prediction period in which the prediction accuracy of the tentative predicted value C is high, or is expected to be high. The future time of about six hours to 48 hours (two days) far from the current time is a prediction period in which the prediction accuracy of the tentative predicted value W is high, or is expected to be high.

[0099] From the current time to a predetermined future time, the blending ratio function of the provisional prediction value T based on the actual value may be set to be dominant. For example, the calculation unit 35 may apply a blending ratio function α, which is a monotonically decreasing function, to the provisional prediction value T. The predetermined future time may be between 30 minutes and 6 hours. The predetermined future time may differ depending on the prediction target. In one example, when the prediction target is the power generated by PV power generation, the predetermined future time may be between 30 minutes and 3 hours. In another example, when the prediction target is power demand prediction, the predetermined future time may be between 3 hours and 24 hours. After the predetermined future time, the blending ratio function of the provisional prediction value W based on weather forecast data may be set to be dominant. For example, the calculation unit 35 may apply a blending ratio function γ, which is a monotonically increasing function, to the provisional prediction value W.

[0100] The calculation unit 35 may select a blend ratio function according to the difference in the prediction period of each prediction method. For example, the calculation unit 35 may set the blend ratio function of a prediction method with a long prediction period among the multiple prediction methods to a monotonically increasing function. The calculation unit 35 may set the blend ratio function of a prediction method with a short prediction period among the multiple prediction methods to a monotonically decreasing function. In one example, assume that there is first weather forecast data with a forecast period of 24 hours and distributed every 30 minutes, and second weather forecast data with a forecast period of 72 hours and distributed every 6 hours. Here, when the desired prediction period is 36 hours, the first weather forecast data does not have a forecast period from 24 hours to 36 hours. The second weather forecast data has a forecast period from 24 hours to 36 hours. However, since the second weather forecast data is a predicted value distributed up to 6 hours ago, it is considered that the prediction accuracy in the most recent prediction period is low. Therefore, the calculation unit 35 applies a blend ratio function of a monotonically decreasing function to the provisional predicted value W1 based on the first weather forecast data. As a result, the provisional forecast value W1 has a higher allocation ratio for the most recent forecast period or the first half of the forecast period. The calculation unit 35 applies a monotonically increasing blending ratio function to the provisional forecast value W2 based on the second weather forecast data. As a result, the allocation ratio for the distant future forecast period or the second half of the forecast period is increased. This makes it possible to improve the prediction accuracy of the predicted value over the entire forecast period. The calculation unit 35 may set the blending ratio function of the forecasting method that has the highest prediction accuracy in the central forecast period when the forecast period is divided into three out of the multiple forecasting methods to a convex function.

[0101] The operation of the EMS 3A will be described with reference to Fig. 10, and an example of an energy prediction method will be described. Fig. 10 is a flowchart showing another example of the energy prediction method as a process flow S2. The process flow S2 differs from the process flow S1 in that it includes steps S16, S17, and S18 instead of step S15. The following describes the process flow S2, focusing mainly on the differences from the process flow S1.

[0102] In step S16, the EMS 3A acquires the third data. For example, the third prediction unit 36 ​​acquires sky image data from the sky photographing device 96 as the third data.

[0103] In step S17, the EMS3A calculates a third tentative predicted value of energy by a third prediction method. For example, the third prediction method may be an energy prediction method using sky image data. The third prediction unit 36 ​​calculates a tentative predicted value C as the third tentative predicted value by the third prediction method using sky image data.

[0104] In step S18, the EMS3A changes the ratio of at least one of the multiple provisional predicted values ​​according to the time of prediction, and blends the multiple provisional predicted values ​​to calculate a predicted value of energy. For example, the calculation unit 35 changes the ratio of at least one of the first provisional predicted value, the second provisional predicted value, and the third provisional predicted value according to the time of prediction, and blends the multiple provisional predicted values ​​to calculate a predicted value B of energy. For example, the calculation unit 35 calculates the predicted value B using equation (13) or the like.

[0105] In the process flow S2, steps S11 and S12, steps S13 and S14, and steps S16 and S17 may be executed in parallel or in series. In relation to steps S12, S14, S17, and S18, the EMS3A may blend prediction errors. For example, in step S12, the first prediction unit 32 may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value W. In step S14, the second prediction unit 34 may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value T. In step S17, the third prediction unit 36 ​​may output a provisional prediction value upper limit and a provisional prediction value lower limit in addition to the provisional prediction value C. In step S18, the calculation unit 35 may further acquire a provisional prediction value upper limit and a provisional prediction value lower limit corresponding to each of the multiple provisional prediction values ​​W, T, and C. The calculation unit 35 may calculate a prediction error range of the energy prediction value B by blending the corresponding tentative prediction value upper limit and tentative prediction value lower limit in accordance with the ratios of the multiple tentative prediction values ​​W, T, and C. For example, the calculation unit 35 may calculate the prediction error range using equation (13). In equation (13), the calculation unit 35 may apply a blending ratio function α to the tentative prediction value upper limit and tentative prediction value lower limit corresponding to the tentative prediction value T. The calculation unit 35 may apply a blending ratio function β to the tentative prediction value upper limit and tentative prediction value lower limit corresponding to the tentative prediction value C. The calculation unit 35 may apply a blending ratio function γ to the tentative prediction value upper limit and tentative prediction value lower limit corresponding to the tentative prediction value W.

[0106] [Hardware configuration] Fig. 11 is a diagram showing an example of a hardware configuration related to the power supply system 1. Fig. 11 shows a computer 100 functioning as the EMS 3, 3A. The computer 100 has a CPU (Central Processing Unit) 101, a main memory unit 102, an auxiliary memory unit 103, a communication control unit 104, an input device 105, and an output device 106. The EMS 3, 3A is composed of one or more computers 100 constituted by these hardware components and software such as programs.

[0107] When the EMS 3, 3A is composed of multiple computers 100, these computers 100 may be connected locally or via a communication network such as the Internet or an intranet. This connection logically constitutes one EMS 3, 3A.

[0108] The CPU 101 executes an operating system, application programs, etc. The main memory 102 is composed of a ROM (Read Only Memory) and a RAM (Random Access Memory). The auxiliary memory 103 is a storage medium composed of a hard disk, a flash memory, etc. The auxiliary memory 103 generally stores a larger amount of data than the main memory 102. The communication control unit 104 is composed of a network card or a wireless communication module. At least a part of the communication function with other devices in the EMS 3, 3A may be realized by the communication control unit 104. The input device 105 is composed of a keyboard, a mouse, a touch panel, a microphone for voice input, etc. The output device 106 is composed of a display, a printer, etc.

[0109] The auxiliary storage unit 103 stores in advance a program 110 (energy prediction program) and data necessary for processing. The program 110 causes the computer 100 to execute each functional element of the EMS 3, 3A. The program 110 causes, for example, processing related to the above-mentioned energy prediction method to be executed in the computer 100. For example, the program 110 is read by the CPU 101 or the main storage unit 102, and causes at least one of the CPU 101, the main storage unit 102, the auxiliary storage unit 103, the communication control unit 104, the input device 105, and the output device 106 to operate. For example, the program 110 reads and writes data in the main storage unit 102 and the auxiliary storage unit 103.

[0110] The program 110 may be provided in a form recorded on a tangible storage medium such as a CD-ROM, a DVD-ROM, a semiconductor memory, etc. The program 110 may be provided as a data signal via a communication network.

[0111] [Effects] An energy prediction device according to one aspect of the present disclosure obtains a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, changes a ratio of at least one of the plurality of tentative prediction values ​​according to the time of prediction, and blends the plurality of tentative prediction values ​​to calculate a predicted value of energy.

[0112] An energy prediction method according to one aspect of the present disclosure includes the steps of obtaining a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, and calculating a predicted value of energy by blending the plurality of tentative prediction values ​​by varying a ratio of at least one of the plurality of tentative prediction values ​​according to a time to be predicted.

[0113] An energy prediction program according to one aspect of the present disclosure causes a computer to execute the steps of obtaining a plurality of tentative prediction values ​​of energy predicted by a plurality of different prediction methods, and calculating a predicted value of energy by changing a ratio of at least one of the plurality of tentative prediction values ​​according to a prediction time and blending the plurality of tentative prediction values.

[0114] In the energy prediction device, the energy prediction method, and the energy prediction program, a predicted value of energy is calculated based on a plurality of provisional predicted values ​​of energy predicted by a plurality of different prediction methods. At this time, the distribution ratio of the plurality of provisional predicted values ​​is changed and blended according to the prediction time. Here, the prediction accuracy of the predicted value of energy may change according to the prediction time. By combining the provisional predicted values ​​by the plurality of different prediction methods according to the prediction time, it is possible to improve the prediction accuracy of the predicted value of energy compared to a single prediction method.

[0115] At least one of the plurality of provisional predicted values ​​is calculated by a prediction method using an actual energy value. The provisional predicted value based on the actual energy value is reflected in the predicted value, thereby improving the prediction accuracy of the predicted energy value.

[0116] A plurality of provisional forecast values ​​are blended so that the ratio of provisional forecast values ​​calculated by a forecasting method using actual energy values ​​is dominant from the current time to a specified future time. The nearest future time can be considered to be a forecast period in which the forecast accuracy of provisional forecast values ​​based on actual energy values ​​is high or is expected to be high. By making the ratio of provisional forecast values ​​based on actual energy values ​​dominant from the current time to a specified future time, the forecast accuracy of the forecasted energy values ​​is further improved.

[0117] The energy is power generated by photovoltaic power generation, and at least one of the provisional predicted values ​​is calculated by a prediction method using weather forecast data. The provisional predicted value based on the weather forecast data is reflected in the predicted value of the power generated by the photovoltaic power generation, thereby improving the prediction accuracy of the predicted value of the energy.

[0118] A plurality of provisional forecast values ​​are blended so that the ratio of provisional forecast values ​​calculated by a forecasting method using weather forecast data becomes dominant from a predetermined future time onward. Future times far from the current time can be considered to be forecast periods in which the prediction accuracy of provisional forecast values ​​based on weather forecast data is high or is expected to be high. By making the ratio of provisional forecast values ​​based on weather forecast data dominant from a future time far from the current time onward, the prediction accuracy of the energy forecast value is further improved.

[0119] For each of the plurality of tentative predicted values, a corresponding tentative predicted value upper limit and a tentative predicted value lower limit are further obtained, and the corresponding tentative predicted value upper limit and tentative predicted value lower limit are blended in accordance with a ratio of the plurality of tentative predicted values ​​to calculate a prediction error range for the predicted value of the energy. By calculating the prediction error range, convenience for a user who uses the predicted value of the energy is improved.

[0120] The prediction methods use different algorithms for prediction. By using different algorithms for each of the prediction methods, it is possible to improve the prediction accuracy of the predicted energy value.

[0121] The prediction methods use different data resources for prediction. The prediction accuracy of the predicted value can reflect the time characteristics of the different data resources. This makes it possible to improve the prediction accuracy of the energy prediction value.

[0122] The prediction methods have different prediction periods. By using different prediction methods for each prediction period, it is possible to improve the prediction accuracy of the predicted energy value. In one example, a tentative predicted value based on a prediction period from the current time to a first future time is combined with a tentative predicted value based on a prediction period from the current time to a second future time after the first future time. In this case, it is possible to improve the prediction accuracy of the predicted energy value from the first future time to the second future time.

[0123] The multiple prediction methods differ from each other in the number of predicted values ​​in the prediction period. By using multiple prediction methods that differ according to the number of predicted values ​​(prediction scores), it is possible to improve the prediction accuracy of the energy predicted value. In one example, a tentative prediction value based on prediction scores for a 30-minute interval and a tentative prediction value based on prediction scores for a 1-hour interval are combined. Since the characteristics due to the time interval are reflected in the tentative prediction value, it is possible to improve the prediction accuracy of the energy predicted value.

[0124] [Variations] The present disclosure is not necessarily limited to the above-described embodiment, and various modifications are possible without departing from the spirit of the present disclosure. For example, the computational resource may be a cloud. The EMS 3 may obtain a plurality of tentative forecast values ​​from an external device. The photovoltaic power generation facility 21 may not be a facility within the microgrid 2.

[0125] In the above embodiment, the blending ratio function α is described as an exponential function, but is not limited thereto. For example, the blending ratio function α may be a power function or a map function. The map function is a function obtained by linearly interpolating a plurality of points of x and y set by a user. In the formula (10), the blending ratio function α may be a monotonically decreasing function based on the characteristics of the tentative predicted value T. Similarly, the design related to (1-α) in the formula (10) may be a monotonically increasing function.

[0126] In the above embodiment, the blending ratio function α is set by the formula (11), but is not limited to this. For example, the blending ratio function α may be set so that the difference (error) between the tentative predicted value T and the actual value and the difference (error) between the tentative predicted value W and the actual value are minimized for each elapsed time from the predicted time. In this case, the error may be minimized under the constraint that the blending ratio function α is a monotonically decreasing function. The error is minimized by the absolute value or the square of the difference between the predicted value and the actual value.

[0127] In the above embodiment, the PV power generation is calculated from the weather forecast data using data such as the rating of the PV array, but this is not limited to the above. For example, the calculation unit 35 may calculate the tentative predicted value W using a statistical model obtained by machine learning the relationship between the past solar radiation amount and temperature, and the PV power generation.

[0128] In the above embodiment, the AR model has been described as an example of the second prediction method, but the present invention is not limited to this. For example, the second prediction method may be a prediction method based on a mathematical model (machine learning model or deep learning model) such as a moving average (MA) model, an auto regressive moving average (ARMA) model, an autoregressive integrated moving average (ARIMA) model, a Holt-Winters model, a gradient boosting decision tree (GBDT) model, a long-short-term memory (LSTM) model, or a gated recurrent unit (GRU) model. In addition, the second prediction method may be a prediction method using, for example, a Kalman filter instead of the minimum AIC model.

[0129] In the above embodiment, the power generated by PV power generation has been described as an example of a prediction target, but the prediction target is not limited to this. For example, the prediction target may be power generated by renewable energy such as wind power generation. Renewable energy generally varies depending on weather conditions, so predictions are often made using weather forecast data. The prediction target may be energy demand such as electricity demand, steam demand, or hydrogen demand. Energy demand generally has a high correlation with weather data such as temperature, so predictions are often made using weather forecast data.

[0130] In the above embodiment, the actual value of the generated power has been described as an example of the actual value, but the present invention is not limited thereto. For example, the output of a renewable energy power source may be suppressed at the request of a power company or the like. Usually, the predicted value of the generated power is desired when the output is not suppressed. For example, the generated power when the output is not suppressed may be calculated based on the output suppression ratio and the generated power measurement value of the PCS, and the calculated value may be used as the generated power actual value. In this way, the generated power when the output is not suppressed can be predicted. In addition, when the prediction target is the power demand, the power demand is measured as a value that is increased or decreased from the original value due to the influence of the demand response. However, when the power demand to be predicted is the original power demand when the demand response is not performed, the baseline value of the demand response may be used as the actual value of the power demand. As a method of calculating the baseline, a method such as "High 4 of 5" is known. The baseline calculation may be performed by the EMS3 or the resource aggregator.

[0131] When the actual value cannot be measured directly, a calculated value or an estimated value may be used as the actual value. For example, when the prediction target is power demand and there are multiple power consumers, it becomes difficult to measure accurately due to the large number of measurement points. The total power consumption may be estimated by artificial intelligence technology or statistical processing based on the values ​​measured for the power consumption of only some of the consumers. The estimated value may be used as the actual demand value of the power consumers.

[0132] The EMS 3 may determine weather fluctuations based on weather forecast data and change the allocation ratio. For example, the calculation unit 35 may determine whether the weather will worsen in a predetermined future time range (e.g., one hour to three hours later). If the calculation unit 35 determines that the weather will worsen, it may increase the ratio of the provisional predicted value W based on the weather forecast data in the predetermined future time range and blend multiple provisional predicted values. This makes it possible to increase the prediction accuracy of the predicted value even when the weather will worsen at a future time close to the current time.

[0133] [Note] This invention is a technology that improves the accuracy of forecasting renewable energy and electricity demand, and by using this technology, it is possible to operate electric power equipment more efficiently by making the most effective use of renewable energy. Therefore, this invention also contributes to the following goals of the Sustainable Development Goals (SDGs) led by the United Nations: Goal 7: Ensure access to affordable, reliable, sustainable and modern energy for all

[0134] The following is a summary of the present disclosure. [1] obtaining a plurality of tentative energy prediction values ​​that are predicted by a plurality of different prediction methods; calculating the predicted value of the energy by varying a ratio of at least one of the plurality of provisional predicted values ​​according to a time to be predicted and blending the plurality of provisional predicted values; Energy forecasting device. [2] The energy prediction device according to [1], wherein at least one of the plurality of tentative predicted values ​​is calculated by a prediction method using an actual value of the energy. [3] The energy prediction device according to [2], wherein the multiple provisional prediction values ​​are blended so that a ratio of the provisional prediction values ​​calculated by a prediction method using the actual energy values ​​becomes dominant from the current time to a specified future time. [4] The energy is solar power generation electricity, The energy prediction device according to any one of [1] to [3], wherein at least one of the plurality of provisional predicted values ​​is calculated by a prediction method using weather forecast data. [5] The energy prediction device according to [4], wherein the multiple provisional forecast values ​​are blended so that a ratio of provisional forecast values ​​calculated by a prediction method using the weather forecast data becomes dominant from a specified future time onwards. [6] further obtaining a corresponding provisional prediction upper limit and a provisional prediction lower limit for each of the plurality of provisional prediction values; calculating a prediction error band of the predicted value of the energy by blending the corresponding tentative prediction value upper limit and the tentative prediction value lower limit according to a ratio of the plurality of tentative prediction values; The energy prediction device according to any one of [1] to [5]. [7] The energy prediction device according to any one of [1] to [6], wherein the plurality of prediction methods use different algorithms for prediction. [8] The energy prediction device according to any one of [1] to [7], wherein the plurality of prediction methods use different data resources for prediction. [9] The energy prediction device according to any one of [1] to [8], wherein the prediction methods have mutually different prediction periods.

[10] The energy prediction device according to any one of [1] to [9], wherein the plurality of prediction methods differ from one another in the number of predicted values ​​in a prediction period.

[11] obtaining a plurality of tentative energy prediction values ​​according to a plurality of different prediction techniques; calculating a predicted value of the energy by varying a ratio of at least one of the plurality of provisional predicted values ​​according to a time to be predicted and blending the plurality of provisional predicted values; The energy forecasting method includes:

[12] obtaining a plurality of tentative energy prediction values ​​according to a plurality of different prediction techniques; calculating a predicted value of the energy by varying a ratio of at least one of the plurality of provisional predicted values ​​according to a time to be predicted and blending the plurality of provisional predicted values; An energy prediction program that causes a computer to run the above. [Explanation of symbols]

[0135] 1. Power supply system 2. Microgrid 3. 3A EMS 21 Solar power generation equipment 22 Power generation measurement section 23 Battery Storage System 24 Connection 25 Receiving power measurement section 26 Transmission power measurement section 31 External Communications Department 32 First Prediction Section 33 Performance Database 34 Second Prediction Section 35 Calculation section 36 Third Prediction Section 90 Power system 95 Weather Forecast Service 96 Aerial photography equipment 101 CPU 102 Main memory 103 Auxiliary storage 104 Communication control section 105 Input Device 106 Output Device 110 Programs B Predicted value C, T, W tentative predictions

Claims

1. obtaining a plurality of tentative energy prediction values ​​that are predicted by a plurality of different prediction methods; Calculating the predicted value of the energy by blending the multiple provisional predicted values ​​by changing a ratio of at least one of the multiple provisional predicted values ​​so as to monotonically increase or decrease depending on the time to be predicted. Energy forecasting device.

2. The energy prediction device according to claim 1 , wherein at least one of the plurality of provisional predicted values ​​is calculated by a prediction method using an actual value of the energy.

3. The energy prediction device according to claim 2 , wherein the plurality of tentative predicted values ​​are blended so that a ratio of the tentative predicted values ​​calculated by the prediction method using the actual energy values ​​becomes dominant from the current time to a predetermined future time.

4. The energy is solar power generation electricity, The energy prediction device according to claim 1 , wherein at least one of the plurality of provisional predicted values ​​is calculated by a prediction method using weather forecast data.

5. The energy prediction device according to claim 4 , wherein the plurality of provisional predicted values ​​are blended so that a ratio of the provisional predicted values ​​calculated by the prediction method using the weather forecast data becomes dominant after a predetermined future time.

6. further obtaining a corresponding provisional prediction upper limit and a provisional prediction lower limit for each of the plurality of provisional prediction values; calculating a prediction error band of the predicted value of the energy by blending the corresponding tentative prediction value upper limit and the tentative prediction value lower limit according to a ratio of the plurality of tentative prediction values; The energy prediction device according to claim 1 .

7. The energy prediction device according to claim 1 , wherein the plurality of prediction methods use different algorithms for prediction.

8. The energy prediction device according to claim 1 , wherein the plurality of prediction methods use different data resources for prediction.

9. The energy prediction device according to claim 1 , wherein the plurality of prediction methods have different prediction periods.

10. The energy prediction device according to claim 1 , wherein the plurality of prediction methods are different from each other in the number of predicted values ​​in a prediction period.

11. obtaining a plurality of tentative energy prediction values ​​according to a plurality of different prediction techniques; calculating a predicted value of the energy by blending the plurality of provisional predicted values ​​by changing a ratio of at least one of the plurality of provisional predicted values ​​so as to monotonically increase or decrease according to a time to be predicted; The energy forecasting method includes:

12. obtaining a plurality of tentative energy prediction values ​​according to a plurality of different prediction techniques; calculating a predicted value of the energy by blending the plurality of provisional predicted values ​​by changing a ratio of at least one of the plurality of provisional predicted values ​​so as to monotonically increase or decrease according to a time to be predicted; An energy prediction program that causes a computer to run the above.

Images