Cloud cover forecasting device, power generation amount forecasting device, cloud cover forecasting method, and power generation amount forecasting method

Abstract

A cloud cover forecasting device (10) comprises: a training unit (11) that obtains a correlation between first time-series data of a cloud shielding rate derived on the basis of a sky image, which is obtained by photographing the sky from an observation location, and second time-series data of cloud shielding rates at a plurality of locations derived on the basis of a satellite image, and that accumulates, as a time difference map and for each of the plurality of locations, a correspondence relationship between the location and a time difference between time-series data for which the correlation is high; and a forecasting unit (12) that uses a preset time difference and a cloud movement direction acquired using the satellite image to derive the cloud shielding rate at a location selected with reference to the time difference map accumulated by the training unit, and outputs a cloud shielding rate forecast value.

Description

Cloud cover forecasting device, power generation forecasting device, cloud cover forecasting method, and power generation forecasting method

[0001] The disclosed technology relates to cloud cover prediction technology.

[0002] Cloud cover prediction technology is a technology for predicting cloud cover related to solar radiation, and is used, for example, in power generation prediction systems that predict the amount of power generated by solar power plants, etc. Conventional power generation prediction systems distribute solar radiation sensors made up of pyranometers and omnidirectional cameras that capture images of the sky at intervals, and predict changes in areas affected by cloud shadows from the solar radiation conditions at two or more locations derived based on the difference in solar radiation intensity measured by each solar radiation sensor, and the direction of cloud movement analyzed from changes in cloud images (Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2022-029818

[0004] Conventional power generation prediction systems have had to install solar radiation sensors, each consisting of a pyranometer, at multiple locations, which has the drawback of making the system configuration large-scale.

[0005] The present disclosure is intended to solve the above problem, and aims to enable cloud cover forecast results to be derived using a smaller configuration than conventional ones.

[0006] The cloud cover forecasting device of the present disclosure is characterized by comprising: a learning unit that calculates a correlation between first time series data of cloud cover ratio derived based on a sky image, which is an image of the sky photographed from an observation point, and second time series data of cloud cover ratio at a plurality of points derived based on satellite images, and accumulates, for each of the plurality of points, a correspondence relationship between the point and the time difference between the time series data with high correlation as a time difference map; and a prediction unit that uses the cloud movement direction obtained using the satellite images and a preset time difference to derive the cloud cover ratio of a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud cover ratio.

[0007] According to the present disclosure, it is possible to derive a cloud cover forecast result using a smaller configuration than conventional ones.

[0008] FIG. 1 is a diagram illustrating an example of a basic configuration of a cloud cover prediction device 10 according to a first embodiment of the present disclosure. FIG. 2 is a diagram illustrating an example of a configuration of a power generation amount prediction system 1A including a cloud cover prediction device 10A according to the first embodiment of the present disclosure. FIG. 3 is a flowchart illustrating an example of a process performed by the cloud cover prediction device 10A according to the first embodiment of the present disclosure. FIG. 4 is a flowchart illustrating an example of a process performed by a power generation amount prediction device 40 according to the first embodiment of the present disclosure. FIG. 5 is a diagram illustrating an example of a configuration of a power generation amount prediction system 1B including a cloud cover prediction device 10B according to a second embodiment of the present disclosure. FIG. 6 is a flowchart illustrating a detailed example of a learning process performed by a learning unit 11B of the cloud cover prediction device 10B according to the second embodiment of the present disclosure. FIG. 7 is a flowchart illustrating an example of a process performed by a cloud coverage ratio derivation unit in the cloud cover prediction device 10B according to the second embodiment of the present disclosure. FIG. 8 is a schematic diagram illustrating a correlation calculation with point i. FIG. 9 is a diagram illustrating an example of a time difference map. FIG. 10 is a flowchart illustrating a detailed example of a prediction process performed by a prediction unit 12B of the cloud cover prediction device 10B according to the second embodiment of the present disclosure. Fig. 11 is a diagram illustrating a configuration example of a power generation amount prediction system 1C including a cloud cover prediction device 10C according to a third embodiment of the present disclosure. Fig. 12 is a flowchart illustrating a detailed example of a prediction process by a prediction unit 12C of the cloud cover prediction device 10C according to the third embodiment of the present disclosure. Fig. 13 is a diagram illustrating a first example of a hardware configuration for realizing functions according to the configuration of the present disclosure. Fig. 14 is a diagram illustrating a second example of a hardware configuration for realizing functions according to the configuration of the present disclosure.

[0009] In order to explain the present disclosure in more detail, embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0010] First Embodiment In the first embodiment, a basic form of the present disclosure will be described.

[0011] An example configuration of a cloud cover prediction device according to a first embodiment of the present disclosure will be described. FIG. 1 is a diagram illustrating an example basic configuration of a cloud cover prediction device 10 according to the first embodiment of the present disclosure. The cloud cover prediction device 10 is a device that predicts cloud cover. Cloud cover is expressed as the percentage of the entire sky at an observation point that is covered by clouds. In the embodiment of the present disclosure, cloud cover is expressed as a "cloud obscuration ratio." The cloud cover prediction device 10 uses the cloud obscuration ratio based on a sky image captured at the observation point and the cloud obscuration ratios at multiple points based on satellite images to accumulate a correspondence relationship between a point in the satellite image and a time difference based on the observation time at the observation point, and predicts the cloud obscuration ratio at the observation point after a preset time has elapsed using the accumulated correspondence relationship and the cloud movement direction. The "time difference setting value," which is the preset time difference, only needs to be set before executing the process of predicting the cloud obscuration ratio, and is received and appropriately set by the cloud cover prediction device 10 before executing the prediction process. The cloud cover prediction device 10 predicts the cloud cover rate at an observation point assuming that a time based on the set "time difference setting value" has elapsed. The cloud cover prediction device 10 shown in Figure 1 is configured to include a learning unit 11 and a prediction unit 12.

[0012] The learning unit 11 has a function of executing a learning process. In the learning process, the learning unit 11 correlates first time-series data of cloud cover ratios derived based on sky images, which are images of the sky captured from an observation point, with second time-series data of cloud cover ratios at multiple points derived based on satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the highly correlated time-series data as a time difference map. Specifically, the learning unit 11 acquires sky images captured at the observation point and derives time-series data of cloud cover ratios using the acquired sky images. This time-series data is also referred to as "first time-series data" in this description. The learning unit 11 also acquires satellite images captured by an artificial satellite and derives time-series data of cloud cover ratios at each of multiple points included in the satellite images using the acquired satellite images. This time-series data is also referred to as "second time-series data" in this description. The learning unit 11 obtains a correlation between the first time series data and the second time series data, and accumulates the correlation as a time difference map indicating the correspondence between the point and the time difference between time series data that are highly correlated for each point among multiple points.

[0013] The prediction unit 12 has a function of executing a prediction process. In the prediction process, the prediction unit 12 uses the cloud movement direction acquired using satellite images and a preset time difference to derive the cloud obscuration ratio of a selected point by referring to a time difference map accumulated by the learning unit, and outputs a predicted value of the cloud obscuration ratio. Specifically, the prediction unit 12 acquires satellite images and derives the cloud movement direction using the acquired satellite images. The cloud movement direction indicates the direction in which clouds move, and can be obtained by determining the movement of clouds in multiple frames of images.

[0014] In addition to the above components, the cloud cover prediction device 10 also includes a control unit (not shown), a storage unit (not shown), and a communication unit (not shown). The control unit (not shown) controls the entire cloud cover prediction device 10 and each of its components. The control unit (not shown), for example, starts up the cloud cover prediction device 10 in response to an external command. The control unit (not shown) also controls the state of the cloud cover prediction device 10 (operating state, such as startup, shutdown, or sleep). The control unit (not shown) also commands each component of the cloud cover prediction device 10 to input and output data. The control unit (not shown) also commands the storage unit (not shown) to store and read data in response to requests from each component of the cloud cover prediction device 10. The storage unit (not shown) stores each piece of data used in the cloud cover prediction device 10. The storage unit (not shown), for example, stores output (output data) from each component of the cloud cover prediction device 10 and outputs data requested by each component to the requesting component. The communication unit (not shown) communicates with external devices. For example, communication is performed between the cloud cover prediction device 10 (100A) and a peripheral device (for example, a camera, a satellite data distribution system, or a power generation prediction device, which will be described later). For example, if the cloud cover prediction device 10 and the peripheral device are not connected by wire, a communication unit (not shown) has a function to perform communication between the cloud cover prediction device 10 and the peripheral device. In addition, the communication unit (not shown) has a function to perform communication with a server device, which is an external device. The control unit (not shown), the memory unit (not shown), and the communication unit (not shown) each have the same functions in the embodiments described later.

[0015] Next, a configuration example of a power generation forecasting system including a cloud cover forecasting device will be described when the cloud cover forecasting device according to the present embodiment is applied to the power generation forecasting system. FIG. 2 is a diagram showing a configuration example of a power generation forecasting system 1A including a cloud cover forecasting device 10A according to the first embodiment of the present disclosure. The power generation forecasting system 1A (1) is a system including multiple devices for forecasting the amount of power generated by photovoltaic power generation. The power generation forecasting system 1A is, for example, a terrestrial system installed on the ground. The power generation forecasting system 1A shown in FIG. 2 is configured to include a cloud cover forecasting device 10A (10), a camera 20, a satellite data distribution system 30, and a power generation forecasting device 40.

[0016] The camera 20 is an imaging device installed at an observation point. The camera 20 acquires a sky image, which is an image of the hemispherical sky centered on the zenith at the observation point. If the observation wavelength of the camera 20 is visible light, it will measure reflected sunlight, and clouds will appear white. The camera 20 outputs the sky image to the cloud cover forecasting device 10A.

[0017] The satellite data distribution system 30 is a system that distributes satellite images acquired from meteorological satellites, etc. The meteorological satellites are, for example, Himawari 8 / 9. When the meteorological satellites are Himawari 8 / 9, the satellite data distribution system 30 can distribute images in the visible band with a terrestrial spatial resolution of 0.5 to 1.0 km. The satellite data distribution system 30 outputs the satellite images to the cloud cover prediction device 10A.

[0018] The cloud cover prediction device 10A predicts the cloud coverage rate in the same manner as the cloud cover prediction device 10 already described. The cloud cover prediction device 10A shown in Fig. 2 is communicably connected to a camera 20 and a satellite data distribution system 30. The cloud cover prediction device 10A acquires sky images from the camera 20. The cloud cover prediction device 10A acquires satellite images from the satellite data distribution system 30. The cloud cover prediction device 10A shown in Fig. 2 is configured to include a learning unit 11A and a prediction unit 12A.

[0019] The learning unit 11A acquires sky images taken at an observation point from the camera 20. The learning unit 11A acquires satellite images from the satellite data distribution system 30. Like the learning unit 11 already described, the learning unit 11A is configured to have the function of executing learning processing.

[0020] The prediction unit 12A acquires satellite images from the satellite data distribution system 30. The prediction unit 12A is configured to have a function of executing prediction processing, similar to the prediction unit 12 already described. The prediction unit 12A outputs a predicted value of the cloud coverage rate to the power generation prediction device 40.

[0021] The power generation prediction device 40 predicts the power generation amount using the cloud cover rate. The power generation prediction device 40 outputs a predicted power generation amount value as a prediction result. The power generation prediction device 40 shown in Fig. 2 predicts the power generation amount by converting the predicted cloud cover rate output by the cloud cover prediction device 10A into a power generation amount, and outputs the predicted power generation amount value as a prediction result.

[0022] In the description, the cloud cover prediction device 10A and the power generation amount prediction device 40 are shown as separate devices, but they may be integrated into one device.

[0023] Next, an example of processing by the cloud cover prediction device will be described. The processing by the cloud cover prediction device 10 shown in FIG. 1 and the processing by the cloud cover prediction device 10A shown in FIG. 2 differ in that the cloud cover prediction device 10A shown in FIG. 2 can specify the source of information used in the processing and the output destination. However, since the internal processing is similar, an example of processing by the cloud cover prediction device 10A will be described here as a representative. FIG. 3 is a flowchart showing an example of processing by the cloud cover prediction device 10A according to the first embodiment of the present disclosure. The processing shown in FIG. 3 is a cloud cover prediction method by the cloud cover prediction device. The cloud cover prediction method may be included in a power generation prediction method by a power generation prediction device. For example, by causing a computer to execute this cloud cover prediction method by a program, the computer can function as a cloud cover prediction device. The cloud cover prediction device 10A shown in FIG. 2 starts the processing shown in FIG. 3 when it receives information from an external or internal control unit (not shown). Specifically, for example, the cloud cover prediction device 10A starts the processing shown in FIG. 3 when it receives a cloud cover prediction command or a power generation prediction command from the outside. (Start)

[0024] The cloud cover prediction device 10A then executes a learning process. (Step ST1110) In the learning process, the learning unit 11A of the cloud cover prediction device 10A executes the learning process. In the learning process, the learning unit 11A correlates first time-series data of cloud cover ratios derived based on sky images, which are images of the sky captured from an observation point, with second time-series data of cloud cover ratios at multiple points derived based on satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the points and the time differences between the highly correlated time-series data as a time difference map. Specifically, the learning unit 11 acquires sky images captured at the observation point and derives time-series data of cloud cover ratios using the acquired sky images. This time-series data is also referred to as "first time-series data" in the description. The learning unit 11 also acquires satellite images captured by an artificial satellite and derives time-series data of cloud cover ratios at each of multiple points included in the satellite images using the acquired satellite images. This time series data is also referred to as “second time series data” in the description. The learning unit 11 obtains a correlation between the first time series data and the second time series data, and accumulates a correspondence relationship between points among a plurality of points and time differences with high correlation as a time difference map.

[0025] The cloud cover prediction device 10A then executes a prediction process (step ST1120). In the prediction process, the prediction unit 12A of the cloud cover prediction device 10A derives the cloud cover ratio for the selected point by using the cloud movement direction acquired using the satellite image and a preset time difference, and outputs a predicted value of the cloud cover ratio by referring to the time difference map accumulated by the learning unit.

[0026] The cloud cover prediction device 10A then executes a prediction result output process (Step ST1130). In the prediction result output process, the prediction unit 12A of the cloud cover prediction device 10A outputs the cloud cover rate prediction value to the power generation amount prediction device 40.

[0027] When the prediction unit 12A outputs the cloud cover rate as the predicted result in the prediction result output process of step ST1130, the cloud cover prediction device 10A then executes an end determination process (step ST1140 "End?"). In the end determination process, a control unit (not shown) of the cloud cover prediction device 10A determines whether to end the processing of the cloud cover prediction device 10A. The control unit (not shown) determines whether to end the processing of the cloud cover prediction device 10A in accordance with an external end command or an execution program. If the control unit (not shown) determines not to end the processing of the cloud cover prediction device 10A (step ST1140 "End?" "NO"), the process returns to step ST1110 and continues. If the control unit (not shown) determines to end the processing of the cloud cover prediction device 10A (step ST1140 "End?" "YES"), the cloud cover prediction device 10A ends the process shown in FIG. 3.

[0028] Next, an example of processing by the power generation prediction device will be described. FIG. 4 is a flowchart showing an example of processing by the power generation prediction device 40 according to the first embodiment of the present disclosure. The processing shown in FIG. 4 is a method for predicting power generation by the power generation prediction device. For example, by having a computer execute this power generation prediction method using a program, the computer can function as the power generation prediction device. The power generation prediction device 40 shown in FIG. 2 starts the processing shown in FIG. 4 when it receives information from an external or internal control unit (not shown). Specifically, for example, the power generation prediction device 40 starts the processing shown in FIG. 4 when it receives a power generation prediction command from the outside. (Start)

[0029] The power generation amount prediction device 40 then executes a cloud cover prediction result acquisition process (step ST1210). In the cloud cover prediction result acquisition process, the power generation amount prediction device 40 acquires the cloud cover rate prediction value, which is the prediction result output by the cloud cover prediction device 10A.

[0030] The power generation amount prediction device 40 then executes a process of inputting a cloud cover forecast result (Step ST1220). In the process of inputting a cloud cover forecast result, the power generation amount prediction device 40 inputs, for example, the acquired cloud cover rate forecast value into the learning model.

[0031] The power generation amount prediction device 40 then executes a power generation amount prediction process (Step ST1230) In the power generation amount prediction process, the power generation amount prediction device 40 acquires, for example, a power generation amount prediction value output from the learning model.

[0032] The power generation amount prediction device 40 then executes a prediction result output process (Step ST1240) In the prediction result output process, the power generation amount prediction device 40 outputs the acquired power generation amount prediction value.

[0033] After executing the prediction result output process in step ST1240, the power generation amount prediction device 40 then executes an end determination process (step ST1250 "End?"). In the end determination process, a control unit (not shown) of the power generation amount prediction device 40 determines whether to end the processing of the power generation amount prediction device 40. The control unit (not shown) determines whether to end the processing of the power generation amount prediction device 40 in accordance with an external end command or an execution program. If the control unit (not shown) determines not to end the processing of the power generation amount prediction device 40 (step ST1250 "NO"), the process returns to step ST1210 and continues. If the control unit (not shown) determines to end the processing of the power generation amount prediction device 40 (step ST1250 "YES"), the power generation amount prediction device 40 ends the processing shown in FIG. 4. (End)

[0034] This embodiment provides an example of the following configuration: a cloud cover forecasting device comprising: a learning unit that calculates a correlation between first time-series data of cloud cover ratios derived from sky images, which are images of the sky captured from an observation point, and second time-series data of cloud cover ratios at multiple points derived from satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the highly correlated time-series data as a time difference map; and a prediction unit that uses a cloud movement direction acquired using satellite images and a preset time difference to derive a cloud cover ratio for a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted cloud cover ratio value. This provides an advantageous effect of providing a cloud cover forecasting device that can derive a cloud cover forecast result with a smaller configuration than conventional devices. Furthermore, a power generation forecasting device including the configuration of a cloud cover forecasting device can achieve the same advantageous effect.

[0035] This embodiment shows an example of the following configuration: A power generation prediction device, comprising: inputting a predicted value of cloud cover rate output by the cloud amount prediction device into a learning model that predicts power generation using the cloud cover rate as an input, and obtaining a predicted power generation amount output by the learning model. As a result, the present disclosure has an effect of providing a power generation prediction device that makes it possible to obtain a power generation amount prediction result using a cloud amount prediction result derived with a smaller configuration than conventional devices.

[0036] This embodiment provides an example of the following configuration: a power generation prediction device comprising: a learning unit that calculates a correlation between first time-series data of cloud cover ratios derived based on sky images, which are images of the sky captured from an observation point, and second time-series data of cloud cover ratios at multiple points derived based on satellite images, and accumulates a time difference map indicating, for each of the multiple points, a correspondence relationship between the point and the time difference between the time-series data with high correlation; a prediction unit that uses a cloud movement direction acquired using the satellite images and a previously accepted time difference to derive a cloud cover ratio for a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted cloud cover ratio; and a power generation prediction value acquisition unit that inputs the predicted cloud cover ratio to a learning model that predicts power generation using the cloud cover ratio as an input, and acquires a predicted power generation value output by the learning model. This disclosure thereby provides an effect of providing a power generation prediction device that enables deriving a cloud amount prediction result with a smaller configuration than conventional devices.

[0037] This embodiment discloses an exemplary embodiment having the following configuration: A cloud amount forecasting method using a cloud amount forecasting device, wherein a learning unit of the cloud amount forecasting device: calculates a correlation between first time-series data of cloud cover ratios derived based on sky images, which are images of the sky captured from an observation point, and second time-series data of cloud cover ratios at multiple points derived based on satellite images; and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the highly correlated time-series data as a time difference map; and a prediction unit of the cloud amount forecasting device: derives a cloud cover ratio for a selected point by referring to the time difference map accumulated by the learning unit using a cloud movement direction acquired using the satellite images and a preset time difference, and outputs a predicted cloud cover ratio value. This provides an advantageous effect of providing a cloud amount forecasting method that enables deriving a cloud amount forecast result with a smaller configuration than conventional methods. Furthermore, a power generation forecasting method including the cloud amount forecasting method achieves the same advantageous effect as described above.

[0038] This embodiment discloses an example configuration as follows: A program causing a computer to operate as a cloud cover forecasting device including: a learning unit that calculates a correlation between first time-series data of cloud cover ratios derived based on sky images, which are images of the sky photographed from an observation point, and second time-series data of cloud cover ratios at multiple points derived based on satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the time-series data with high correlation as a time difference map; and a prediction unit that uses a cloud movement direction acquired using the satellite images and a preset time difference to derive a cloud cover ratio for a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud cover ratio. This provides an effect of providing a program that enables deriving a cloud cover forecast result with a smaller configuration than conventional programs.

[0039] This embodiment shows an example of the following configuration: A power generation prediction method using a power generation prediction device, wherein a learning unit of the power generation prediction device calculates a correlation between time-series data of cloud blockage ratios derived based on sky images, which are images of the sky photographed from an observation point, and time-series data of cloud blockage ratios at multiple points derived from satellite images, and accumulates, as a time difference map, correspondence relationships of time differences that have a high correlation with points in the satellite images, a prediction unit of the power generation prediction device uses the cloud movement direction acquired from the satellite images and a preset time difference to derive the cloud blockage ratio of a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud blockage ratio, and a power generation prediction value acquisition unit of the power generation prediction device inputs the predicted value of the cloud blockage ratio output by the prediction unit to a learning model that predicts power generation using the cloud blockage ratio as an input, and acquires a predicted power generation value that is a prediction result output by the learning model. As a result, the present disclosure has the effect of providing a power generation forecasting method that makes it possible to obtain a power generation forecast result using a cloud cover forecast result derived using a smaller-scale configuration than conventional methods.

[0040] Embodiment 2. In embodiment 2, a more detailed configuration example of the configuration according to embodiment 1 will be described. In embodiment 2, among the components according to embodiment 2, components similar to the components according to embodiment 1 already described will be given the same names and similar reference numerals (with some modifications), and duplicated descriptions will be omitted as appropriate.

[0041] A configuration example of a cloud cover prediction device and a system including the device according to a second embodiment of the present disclosure will be described. Fig. 5 is a diagram showing a configuration example of a power generation amount prediction system 1B including a cloud cover prediction device 10B according to the second embodiment of the present disclosure. The power generation amount prediction system 1B shown in Fig. 5 is configured to include a cloud cover prediction device 10B (10), a camera 20, a satellite data distribution system 30, and a power generation amount prediction device 40. The camera 20 and the satellite data distribution system 30 are similar to the camera 20 and satellite data distribution system 30 already described, and detailed description thereof will be omitted.

[0042] The cloud cover prediction device 10B predicts the cloud cover ratio in the same manner as the cloud cover prediction device 10A already described. The cloud cover prediction device 10B shown in Figure 5 includes a learning unit 11B (11) and a prediction unit 12B (12).

[0043] The learning unit 11B is configured to have the function of executing learning processing, similar to the learning unit 11A already described. The learning unit 11B shown in Fig. 5 is configured to include a camera image storage unit 101B (101), a first cloud cover ratio derivation unit 102B (102), a first time-series data generation unit 103B (103), a second cloud cover ratio derivation unit 202B (202), a second time-series data generation unit 203B (203), a correlation calculation unit 301B (301), a time difference map generation unit 302B (302), and a time difference map storage unit 303B (303).

[0044] The camera image storage unit 101B stores sky images. The camera image storage unit 101B stores image signals output by the camera 20 in chronological order.

[0045] The first cloud cover ratio deriving unit 102B derives the cloud cover ratio based on a sky image, which is an image of the sky captured from an observation point. Specifically, for example, the first cloud cover ratio deriving unit 102B derives the cloud cover ratio for a range of a region of interest in the sky image stored in the camera image storage unit 101B.

[0046] The first time-series data generation unit 103B generates time-series data of the cloud coverage ratio derived based on the sky image. Specifically, for example, the first time-series data generation unit 103B converts the cloud coverage ratio derived based on the sky image into time-series data for a certain time range. The certain time range is, for example, a time range from sunrise to sunset (sunset), and is a preset time range. In this description, this time-series data can be referred to as "first time-series data" or "camera cloud coverage ratio time-series data." The first time-series data generation unit 103B outputs the first time-series data (camera cloud coverage ratio time-series data) to the correlation calculation unit 301B.

[0047] The second cloud cover ratio deriving unit 202B derives cloud cover ratios at multiple points based on the satellite image. Specifically, for example, the second cloud cover ratio deriving unit 202B derives cloud cover ratios for ranges in the satellite image that are approximately the same area as the range of the region of interest, for each of multiple points in a range that is correlated with the region of interest from which the cloud cover ratio of the sky image was derived. This ensures that the range of the cloud cover ratio derived from the satellite image and the range of the region of interest in the sky image are not unrelated ranges.

[0048] The second time series data generation unit 203B generates time series data of cloud coverage ratios at multiple locations derived based on satellite images. Specifically, for example, the second time series data generation unit 203B converts the cloud coverage ratios derived based on satellite images into time series data for a certain time range. This certain time range is the same as the time range of the first time series data (camera cloud coverage ratio time series data) already described. In this description, this time series data can be referred to as "second time series data" or "satellite cloud coverage ratio time series data." The second time series data generation unit 203B outputs the second time series data (satellite cloud coverage ratio time series data) to the correlation calculation unit 301B.

[0049] The correlation calculation unit 301B calculates the correlation between the first time series data (camera cloud cover rate time series data) and the second time series data (satellite cloud cover rate time series data). Specifically, for example, the correlation calculation unit 301B calculates the correlation between the first time series data of cloud cover rate derived based on a sky image, which is an image of the sky photographed from an observation point, and the second time series data of cloud cover rate derived for a plurality of points using satellite images. The correlation calculation unit 301B calculates, for example, a cross-correlation coefficient R(τ) between the first time series data and the second time series data.

[0050] The time difference map generation unit 302B calculates the time difference between highly correlated time-series data for each of multiple locations included in the satellite image based on the results of the correlation calculation by the correlation calculation unit 301B. Specifically, for example, the time difference map generation unit 302B calculates the time difference τ=τ0 at which the cross-correlation coefficient R(τ) calculated by the correlation calculation unit 301B is maximized, and generates a time difference map that shows the correspondence between the time range (e.g., date) used in the calculation, the locations in the satellite image, and the time difference τ0 (maximum time difference). The time difference map generation unit 302B stores the generated time difference map in the time difference map storage unit 303B.

[0051] The time difference map storage unit 303B stores the time difference map generated by the time difference map generation unit 302B. Specifically, the time difference map storage unit 303B stores a time difference map indicating the correspondence between each point among multiple points and the time difference between highly correlated time-series data. As a result, the time difference map storage unit 303B accumulates the correspondence between time ranges (e.g., dates), points in satellite images, and time difference τ0 (maximum time difference).

[0052] The prediction unit 12B is configured to have a function of executing prediction processing, similar to the prediction unit 12A already described. The prediction unit 12B shown in Fig. 5 is configured to include a cloud movement direction derivation unit 401B (401), a point selection unit 402B (402), and a third cloud coverage ratio derivation unit 403B (403).

[0053] The cloud movement direction deriving unit 401B derives the cloud movement direction, which is the direction in which the clouds are moving, using the received image. Specifically, for example, the cloud movement direction deriving unit 401B derives the cloud movement direction from multiple frames of current satellite images received from the satellite data distribution system 30.

[0054] The point selection unit 402B refers to the time difference map and selects a point using the cloud movement direction and a preset time difference. The preset time difference is a time difference setting value indicating the time difference between the current time and the time at which the user wants to predict the cloud coverage rate. The time difference setting value indicates the time difference between the current time, such as 15 minutes from the current time, and a time in the near future. Specifically, for example, the point selection unit 402B refers to the time difference map and determines the direction of the point using the cloud movement direction, and further determines the distance from the observation point using the time difference (preset time difference) at which the user wants to predict the power generation amount, thereby selecting a point from among multiple points on the time difference map.

[0055] The third cloud cover ratio deriving unit 403B derives the cloud cover ratio at the point selected by the point selecting unit 402B based on the current cloud image. Specifically, for example, the third cloud cover ratio deriving unit 403B acquires a current cloud image and derives the cloud cover ratio at the selected point using the current cloud image. The current cloud image in this embodiment is a satellite image output by a satellite data distribution system 30 that distributes data from artificial satellites. In this case, the satellite data distribution system 30 can be expressed as a cloud image acquisition means. The third cloud cover ratio deriving unit 403B outputs the derived cloud cover ratio to the power generation amount prediction device 40. The cloud cover ratio output by the third cloud cover ratio deriving unit 403B of the cloud amount prediction device 10B can be expressed as a cloud cover ratio prediction value.

[0056] The power generation prediction device 40 predicts the power generation amount using the cloud cover rate. The power generation prediction device 40 outputs a predicted power generation amount value, which is the prediction result. The power generation prediction device 40 shown in FIG. 5 predicts the power generation amount by converting the predicted cloud cover rate value output by the cloud cover prediction device 10B into a power generation amount, and outputs a predicted power generation amount value, which is the prediction result. The power generation prediction device 40 shown in FIG. 5 is configured to include a power generation amount predicted value acquisition unit 41 and a learning model unit 42.

[0057] The power generation forecast value acquisition unit 41 acquires the cloud cover rate (cloud cover rate forecast value) output by the cloud cover prediction device 10B and inputs the acquired cloud cover rate to the learning model unit 42. The power generation forecast value acquisition unit 41 acquires the power generation amount output by the learning model unit 42 and outputs the acquired power generation amount. In other words, the power generation forecast value acquisition unit 41 acquires the power generation forecast value using the cloud cover rate forecast value.

[0058] The learning model unit 42 includes a learning model that has been constructed by learning in advance so as to input the cloud cover rate and output the power generation amount. That is, the learning model of the learning model unit 42 has been constructed by learning in advance so as to associate the cloud cover rate with the power generation amount. The power generation amount output by the learning model is a power generation amount prediction value. That is, when the cloud cover rate prediction value is input from the power generation amount prediction value acquisition unit 41, the learning model unit 42 converts the cloud cover rate prediction value into a power generation amount prediction value and outputs it.

[0059] Next, an example of processing by the learning unit 11B in the cloud cover prediction device 10B according to the second embodiment of the present disclosure will be described. FIG. 6 is a flowchart illustrating a detailed example of learning processing by the learning unit 11B of the cloud cover prediction device 10B according to the second embodiment of the present disclosure. The processing illustrated in FIG. 6 is included in a cloud cover prediction method performed by a cloud cover prediction device. The cloud cover prediction method may be included in a power generation prediction method performed by a power generation prediction device. For example, by having a computer execute this cloud cover prediction method using a program, the computer can function as a cloud cover prediction device. The learning unit 11B of the cloud cover prediction device 10B shown in FIG. 5 starts the processing illustrated in FIG. 6 upon receiving information from an external or internal control unit (not shown). Specifically, for example, the learning unit 11B of the cloud cover prediction device 10A starts the processing illustrated in FIG. 6 upon receiving a cloud cover prediction command or a power generation prediction command from the outside. (Start)

[0060] The learning unit 11B then executes a process of extracting images from the camera image storage unit 101B. (Step ST2301) In this process, the first cloud coverage ratio derivation unit 102B of the learning unit 11B extracts multiple frames of camera images (sky images) within a certain time range from the camera image storage unit 101B.

[0061] The learning unit 11B then executes a process to derive a cloud cover ratio. (Step ST2302) In this process, the first cloud cover ratio derivation unit 102B of the learning unit 11B derives a cloud cover ratio based on a sky image, which is an image of the sky photographed from an observation point. The first cloud cover ratio derivation unit 102B derives a cloud cover ratio for each received frame of the sky image. The first cloud cover ratio derivation unit 102B outputs the cloud cover ratio to the first time-series data generation unit 103B.

[0062] Here, the derivation process in the process for deriving the cloud coverage factor will be described. FIG. 7 is a flowchart showing an example of the process of the cloud coverage factor derivation unit in the cloud coverage forecasting device 10B according to the second embodiment of the present disclosure. The process shown in FIG. 7 is a cloud coverage forecasting method performed by the cloud coverage forecasting device. The cloud coverage forecasting method may be included in the power generation forecasting method performed by the power generation forecasting device. For example, by having a computer execute this cloud coverage forecasting method using a program, the computer can function as the cloud coverage forecasting device. The cloud coverage factor derivation unit (102B, 202B) shown in FIG. 5 starts the process shown in FIG. 7 upon receiving a command from a control unit (not shown) of the cloud coverage forecasting device 10B, the learning unit 11B, or the prediction unit 12B. Specifically, for example, the cloud coverage factor derivation unit (102B, 202B) starts the process shown in FIG. 7 upon receiving a cloud coverage factor derivation command and a camera image or a satellite image. (Start)

[0063] The cloud coverage ratio deriving unit then performs a process of dividing the image into grid-like regions (step ST2401).

[0064] The cloud coverage ratio deriving unit then performs a process of calculating the average value of pixel values ​​in each region (step ST2402).

[0065] The cloud coverage ratio deriving unit then performs a binarization process using a certain threshold value (step ST2403).

[0066] The cloud obscuration ratio deriving unit then executes a process of dividing the sum of the binarized values ​​by the area to obtain the cloud obscuration ratio (step ST2404). In this process, the cloud obscuration ratio deriving unit determines the value obtained by dividing the sum of the binarized values ​​within the region of interest by the number of regions of interest as the cloud obscuration ratio.

[0067] The cloud coverage ratio deriving unit then ends the processing shown in FIG. 7 (END).

[0068] Returning to the explanation of Fig. 6, the learning unit 11B then executes a process of generating time series data (=first time series data (camera cloud coverage rate time series data) "Ac_cam(t)"). (Step ST2303) In this process, the first time series data generation unit 103B of the learning unit 11B organizes the cloud coverage rate as time series data. The first time series data generation unit 103B outputs the first time series data (camera cloud coverage rate time series data) to the correlation calculation unit 301B.

[0069] The learning unit 11B executes the processes of steps ST2304 to ST2306 in parallel with the processes of steps ST2301 to ST2303. The learning unit 11B executes a process of receiving satellite images. (Step ST2304) In this process, the second cloud coverage ratio derivation unit 202B of the learning unit 11B acquires from the satellite data distribution system 30 a plurality of frames of satellite images covering a range approximately the same as the time range extracted in the process of step ST2301.

[0070] The learning unit 11B then executes a process to derive a cloud coverage ratio. (Step ST2305) In this process, the second cloud coverage ratio derivation unit 202B of the learning unit 11B derives cloud coverage ratios for multiple points using satellite images. Specifically, the second cloud coverage ratio derivation unit 202B derives the cloud coverage ratio in accordance with the process of the flowchart shown in Fig. 7. However, since it is expected that the spatial resolution of satellite images is lower than that of camera images, in this case, the second cloud coverage ratio derivation unit 202B omits the process of step ST2402 if the grid-like region obtained by dividing the image in the process of step ST2401 contains only one pixel.

[0071] The learning unit 11B then executes a process to generate time series data (= second time series data (satellite cloud coverage rate time series data) "Ac_sat_i(t)"). (Step ST2306) In this process, the second time series data generation unit 203B of the learning unit 11B organizes the cloud coverage rate as time series data. The time series data generated from the cloud coverage rate of point i is called Ac_sat_i(t). The second time series data generation unit 203B outputs the second time series data (satellite cloud coverage rate time series data) to the correlation calculation unit 301B.

[0072] After the learning unit 11B has executed the processes from step ST2301 to step ST2303 and the processes from step ST2304 to step ST2306, it then executes a process of calculating correlation. (Step ST2307) Fig. 8 is a schematic diagram for explaining the correlation calculation with point i. Fig. 8 shows an image of the processes of step ST2307 and step ST2308. In this process, the correlation calculation unit 301B of the learning unit 11B calculates the cross-correlation coefficient R(τ) between the camera cloud coverage factor time series data Ac_cam(t) and the satellite cloud coverage factor time series data Ac_sat_i(t) based on equation (1). where μ ｃａｍ , μ ｓａｔ＿ｉ are the average values ​​of the camera cloud obscuration factor time series data and the average values ​​of the satellite cloud obscuration factor time series data at point i, respectively, and are expressed by equations (2) and (3).

[0073] The learning unit 11B then executes a process of determining the time difference τ0 at which the cross-correlation coefficient R(τ) received from the correlation calculation unit 301B is maximized (step ST2308).

[0074] The learning unit 11B then executes a process of adding the time difference τ0 to the time difference map (step ST2309). In this process, the time difference map generation unit 302B of the learning unit 11B adds the time difference τ0 to the time difference map together with the location and the time range (e.g., date) used in the calculation.

[0075] The learning unit 11B then executes a process of storing the new data in the time difference map storage unit (step ST2310). In this process, the time difference map generation unit 302B of the learning unit 11B adds new data to the time difference map stored in the time difference map storage unit 303B and stores the new data.

[0076] Next, the learning unit 11B executes a process of determining whether calculations have been completed for all points (step ST2311). In this process, a control unit (not shown) of the learning unit 11B determines whether the time difference map generating unit 302B has generated time difference maps for all of the predetermined points.

[0077] If the control unit (not shown) of the learning unit 11B determines that there are any incomplete points (step ST2311 "Calculation at all points?" "NO"), it transitions to the processing of step S2312. The learning unit 11B then executes the point change processing. (step ST2312) In the point change processing, the learning unit 11B changes the point and returns to the processing of step S2305.

[0078] If the control unit (not shown) of the learning unit 11B determines that calculations have been performed for all points (step ST2311 "Calculated at all points?" "YES"), it ends the processing shown in FIG. 6 (END).

[0079] 9 is a diagram showing an example of a time difference map 1000. A time difference map 1000 associates time differences between time series data for each time range, such as date, and for each location.

[0080] FIG. 10 is a flowchart showing a detailed example of a prediction process by the prediction unit 12B of the cloud cover prediction device 10B according to the second embodiment of the present disclosure. The process shown in FIG. 10 is a cloud cover prediction method by the cloud cover prediction device. The cloud cover prediction method may be included in a power generation prediction method by a power generation prediction device. For example, by having a computer execute this cloud cover prediction method using a program, the computer can function as a cloud cover prediction device. The cloud cover prediction device 10B shown in FIG. 5 starts the process shown in FIG. 10 upon receiving information from an external or internal control unit (not shown). Specifically, for example, the cloud cover prediction device 10B starts the process shown in FIG. 10 upon receiving a cloud cover prediction command or a power generation prediction command from the outside. (Start)

[0081] The prediction unit 12B then executes a process of receiving satellite images (step ST2501). In this process, the cloud movement direction deriving unit 401B of the prediction unit 12B receives a plurality of frames of images, including images captured at times close to the current time, from the satellite data distribution system 30.

[0082] The prediction unit 12B then executes a process of deriving the cloud movement direction. (Step ST2502) In this process, the cloud movement direction derivation unit 401B of the prediction unit 12B derives the cloud movement direction from the received multiple frames of images. The cloud movement direction derivation unit 401B derives the cloud movement direction using the current multiple frames of satellite images.

[0083] The prediction unit 12B then executes a process of selecting a location (step ST2503). In this process, the location selection unit 402B of the prediction unit 12B refers to the time difference map and selects a location using the cloud movement direction and a preset time difference. Specifically, the location selection unit 402B compares the derived current cloud movement direction with the time difference map stored in the learning unit 11B, and selects a location that has a high correlation with the cloud coverage rate at the time difference (e.g., 15 minutes later) that is to be predicted.

[0084] The prediction unit 12B then executes a process to derive the cloud cover ratio for the selected point. (Step ST2504) In this process, the third cloud cover ratio derivation unit 403B of the prediction unit 12B acquires a current cloud image and derives the cloud cover ratio for the selected point using the current cloud image. Specifically, for example, the third cloud cover ratio derivation unit 403B derives the cloud cover ratio for the point selected by the point selection unit 402B using, of the satellite images received in the process of step S2501, the image captured at the time closest to the current time, and outputs the cloud cover ratio to the power generation amount prediction device 40. The process of deriving the cloud cover ratio is similar to the process described in FIG. 7 , and therefore will not be described again.

[0085] When the third cloud cover ratio deriving unit 403B outputs the cloud cover ratio to the power generation amount prediction device 40, the prediction unit 12B ends the processing shown in FIG. 10 (END).

[0086] As described above, correlation processing is performed using a statistical quantity called cloud obscuration rate without using individual cloud mass information from image data, which reduces the amount of calculation required and enables the system to be built at low cost. In addition, by correlating the cloud obscuration rate, which is directly related to the location and power generation amount, it is possible to determine the time difference, which contributes to improving the accuracy of power generation forecasts.

[0087] The present embodiment further discloses an example of the following configuration: The cloud amount forecasting device, wherein the learning unit comprises: a correlation calculation unit that performs a correlation calculation between the first time series data, which is time series data of cloud cover ratios derived based on sky images, which are images of the sky photographed from an observation point, and the second time series data, which is time series data of cloud cover ratios derived for a plurality of points using satellite images, a time difference map generation unit that uses the results of the correlation calculation to determine the time difference between highly correlated time series data for each of a plurality of points included in the satellite images, and a time difference map storage unit that stores a time difference map that indicates the correspondence between each of the plurality of points and the time difference between highly correlated time series data, and the prediction unit comprises: a cloud movement direction derivation unit that derives a cloud movement direction using current frames of satellite images, a point selection unit that selects a point using the cloud movement direction and a preset time difference with reference to the time difference map, and a cloud cover ratio derivation unit that acquires a current cloud image and derives the cloud cover ratio at the selected point using the current cloud image. As a result, the present disclosure further provides an effect of providing a cloud cover prediction device that enables deriving cloud cover prediction results with a smaller configuration than conventional devices. Furthermore, the present disclosure provides the same effect as the above by applying the above configuration to a system, the above cloud cover prediction method, or the above program. Furthermore, the present disclosure provides the same effect as the above by including the configuration of the cloud cover prediction device in a power generation forecasting device.

[0088] This embodiment further discloses an exemplary embodiment having the following configuration: A cloud cover prediction device, characterized in that the current cloud image is a satellite image output by a satellite data distribution system that distributes data from an artificial satellite. As a result, the present disclosure further provides an effect of providing a cloud cover prediction device that can derive cloud cover prediction results with a smaller configuration than conventional devices, since the current cloud image can be used in combination. Furthermore, the present disclosure achieves the same effect as the above by applying the above configuration to a system, the cloud cover prediction method, or the program. Furthermore, the present disclosure achieves the same effect as the above by including the configuration of the cloud cover prediction device in the power generation prediction device.

[0089] This embodiment further discloses the following exemplary configuration: A power generation prediction device, comprising: inputting a predicted value of cloud cover rate output by the cloud amount prediction device into a learning model that predicts power generation using the cloud cover rate as an input, and obtaining a predicted power generation amount output by the learning model. This provides an effect of providing a power generation prediction device that makes it possible to obtain a power generation amount prediction result using a cloud amount prediction result derived with a smaller configuration than conventional devices. Furthermore, the present disclosure provides the same effect as the above by applying the above configuration to a system, the power generation amount prediction method, or the program.

[0090] Embodiment 3. In the above-described embodiment 2, a satellite image received from a satellite data distribution system is used as input to the third cloud cover ratio derivation unit of the prediction unit in the cloud amount forecasting device. In this embodiment, an embodiment is shown in which the cloud cover ratio is calculated using an image from a camera i installed at a point i selected by a point selection unit, among cameras distributed across various locations. In embodiment 3, components of embodiment 3 that are similar to components of embodiment 1 or embodiment 2 already described are given the same names and similar reference numerals (with some modifications), and redundant explanations are omitted as appropriate.

[0091] A configuration example of a cloud cover prediction device and a system including the device according to a third embodiment of the present disclosure will be described. FIG. 11 is a diagram showing a configuration example of a power generation amount prediction system 1C including a cloud cover prediction device 10C according to the third embodiment of the present disclosure. The power generation amount prediction system 1C (1) shown in FIG. 11 includes a cloud cover prediction device 10C (10), a camera 20, a satellite data distribution system 30, a power generation amount prediction device 40, and a camera i 50 (the camera i 50 will also be referred to as "camera (i) 50" in the description). The camera 20 is configured to have the same functions as the camera 20 already described. The satellite data distribution system 30 is configured to have the functions of the satellite data distribution system 30 already described.

[0092] The camera (i) 50 is installed at each point different from the observation point, and acquires an image of the sky in a hemisphere centered on the zenith at the observation point (sky image), and outputs it to the cloud cover forecasting device 10C.

[0093] The cloud cover forecasting device 10C shown in FIG. 11 includes a learning unit 11C (11) and a forecasting unit 12C (12).

[0094] The learning unit 11C is configured to have the same functions as the learning unit 11B already described. The learning unit 11C shown in Figure 11 is configured to include a camera image storage unit 101C (101), a first cloud coverage ratio derivation unit 102C (102), a first time-series data generation unit 103C (103), a second cloud coverage ratio derivation unit 202C (202), a second time-series data generation unit 203C (203), a correlation calculation unit 301C (301), a time difference map generation unit 302C (302), and a time difference map storage unit 303C (303). Since the configurations are similar to those already described with the same names and similar reference numerals (with some modifications), detailed description will be omitted.

[0095] The prediction unit 12C shown in FIG. 11 includes a cloud movement direction derivation unit 401C (401), a point selection unit 402C (402), and a third cloud coverage ratio derivation unit 403C (403).

[0096] The cloud movement direction deriving unit 401C is configured to derive the cloud movement direction, which is the direction in which clouds are moving, using the received image, similarly to the cloud movement direction deriving unit 401B already described.

[0097] The location selection unit 402C refers to the time difference map and selects a location using the cloud movement direction and a preset time difference. The preset time difference is a time difference setting value indicating the time difference between the current time and the time at which the user wants to predict the cloud coverage rate. The time difference setting value indicates the time difference between the current time, such as 15 minutes from the current time, and a time in the near future. Specifically, for example, the location selection unit 402C refers to the time difference map and determines the direction of the location using the cloud movement direction, and further determines the distance from the observation location using the time difference (preset time difference) at which the user wants to predict the power generation amount, thereby selecting a location from among multiple locations on the time difference map. The location selection unit 402C instructs the camera (i) 50 installed at the selected location (i) to output the captured current cloud image to the cloud cover prediction device 10C.

[0098] The third cloud cover ratio deriving unit 403C of the cloud amount forecasting device 10C derives the cloud cover ratio at a selected point from the acquired current cloud image. The current cloud image is an image taken by a camera (i) 50 installed at multiple points different from the observation point. In this case, the camera (i) 50 can be expressed as a cloud image acquisition means.

[0099] The power generation amount prediction device 40 is configured to function in the same manner as the power generation amount prediction device 40 already described.

[0100] Next, an example of processing by a cloud cover prediction device according to a third embodiment of the present disclosure will be described. FIG. 12 is a flowchart illustrating a detailed example of prediction processing by a prediction unit 12C of a cloud cover prediction device 10C according to a third embodiment of the present disclosure. The processing illustrated in FIG. 12 is a cloud cover prediction method performed by a cloud cover prediction device. The cloud cover prediction method may be included in a power generation prediction method performed by a power generation prediction device. For example, by causing a computer to execute this cloud cover prediction method using a program, the computer can function as a cloud cover prediction device. The cloud cover prediction device 10A illustrated in FIG. 11 starts the processing illustrated in FIG. 12 upon receiving a command from an external or internal control unit (not shown). Specifically, for example, the cloud cover prediction device 10A starts the processing illustrated in FIG. 12 upon receiving a cloud cover prediction command or a power generation prediction command from the outside. (Start) Of the processes shown in Figure 12, the processes of steps ST3501, ST3502, and ST3503 are similar to the processes of steps ST2501, ST2502, and ST2503 in the flowchart of Figure 10 already described, and therefore their explanations are omitted.

[0101] The prediction unit 12C then executes a process of receiving a camera image of the selected point (step ST3504). In this process, the point selection unit 402C of the prediction unit 12C instructs the camera (i) 50 installed at the selected point (i) to output a captured current cloud image to the cloud cover prediction device 10C. The camera (i) 50 outputs the captured current cloud image to the cloud cover prediction device 10C in accordance with the instruction from the point selection unit 402C. The third cloud coverage ratio derivation unit 403C of the prediction unit 12C in the cloud cover prediction device 10C acquires a current cloud image (camera image) from the camera (i) 50.

[0102] The prediction unit 12C then executes a process to derive the cloud cover ratio of the camera image. (Step ST3505) In this process, the third cloud cover ratio derivation unit 403C of the prediction unit 12C derives the cloud cover ratio in the same manner as the process shown in the flowchart of Fig. 7 and outputs a predicted value of the cloud cover ratio. The cloud cover ratio of the current cloud image (camera image) is derived. In this process, the third cloud cover ratio derivation unit 403C of the prediction unit 12B derives the cloud cover ratio for the point selected by the point selection unit 402C and outputs it to the power generation amount prediction device 40. The process of deriving the cloud cover ratio is the same as the process described in Fig. 7, so a description thereof will be omitted.

[0103] When the third cloud cover ratio deriving unit 403C outputs the cloud cover ratio to the power generation amount prediction device 40, the prediction unit 12C ends the processing shown in FIG. 12 (END).

[0104] The configuration according to this embodiment allows the selected point and the cloud coverage rate to correspond more closely, thereby improving the accuracy of the cloud amount forecast results.

[0105] This embodiment further discloses an exemplary embodiment having the following configuration: A cloud cover prediction device, characterized in that the current cloud image is an image captured by cameras installed at multiple locations different from the observation location. This provides an advantageous effect of providing a cloud cover prediction device that enables a cloud cover prediction result to be derived with a smaller configuration than conventional devices and that enables the accuracy of the cloud cover prediction result to be improved. Furthermore, the present disclosure provides an advantageous effect similar to the above-described advantageous effect by applying the above-described configuration to a system, the above-described cloud cover prediction method, or the above-described program. Furthermore, the present disclosure provides an advantageous effect similar to the above-described advantageous effect by including the configuration of the cloud cover prediction device in the power generation prediction device.

[0106] Here, a hardware configuration for realizing the functions of the present disclosure will be described. Fig. 13 is a diagram showing a first example of a hardware configuration for realizing the functions of the configuration of the present disclosure. Fig. 14 is a diagram showing a second example of a hardware configuration for realizing the functions of the configuration of the present disclosure. The cloud cover forecasting devices 10, 10A, 10B, and 10C of the present disclosure are each realized by the hardware shown in Fig. 13 or Fig. 14.

[0107] As shown in FIG. 13 , each of the cloud cover prediction devices 10, 10A, 10B, and 10C and the power generation prediction device 40 is configured with, for example, a processor 10001, a memory 10002, an input / output interface 10003, and a communication circuit 10004. The processor 10001 and the memory 10002 are, for example, installed in a computer. Each function of the present disclosure is realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 10002. That is, the memory 10002 stores the computer including learning units 11, 11A, 11B, 11C, a first cloud coverage ratio deriving unit 102 (102B, 102C), a first time series data generating unit 103 (103B, 103C), a second cloud coverage ratio deriving unit 202 (202B, 202C), a second time series data generating unit 203 (203B, 203C), a correlation calculation unit 301 (301B, 301C), The storage unit stores a time difference map generation unit 302 (302B, 302C), a prediction unit 12, 12A, 12B, 12C, a cloud movement direction derivation unit 401 (401B, 401C), a point selection unit 402 (402B, 402C), a third cloud coverage rate derivation unit 403 (403B, 403C), a power generation prediction value acquisition unit 41, a learning model unit 42, and a program for functioning as a control unit not shown. The processor 10001 reads and executes the program stored in the memory 10002, whereby the learning units 11, 11A, 11B, 11C, the first cloud coverage ratio derivation unit 102 (102B, 102C), the first time series data generation unit 103 (103B, 103C), the second cloud coverage ratio derivation unit 202 (202B, 202C), the second time series data generation unit 203 (203B, 203C), the correlation The functions of the calculation unit 301 (301B, 301C), the time difference map generation unit 302 (302B, 302C), the prediction unit 12, 12A, 12B, 12C, the cloud movement direction derivation unit 401 (401B, 401C), the point selection unit 402 (402B, 402C), the third cloud coverage ratio derivation unit 403 (403B, 403C), the power generation amount forecast value acquisition unit 41, the learning model unit 42, and a control unit (not shown) are realized. The program causes a computer to execute the procedures or methods of each of the above components.The memory 10002 or another memory (not shown) implements the camera image storage unit 101 (101B, 101C), the time difference map storage unit 303 (303B, 303C), and a storage unit (not shown). The communication circuit 10004 implements a communication unit (not shown).

[0108] The processor 10001 is, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a microprocessor, a microcontroller, or a DSP (Digital Signal Processor). The memory 10002 may be a non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable Read Only Memory), or flash memory, or a magnetic disk such as a hard disk or flexible disk, or an optical disk such as a CD (Compact Disc) or DVD (Digital Versatile Disc), or a magneto-optical disk. The processor 10001 and the memory 10002 or the communication circuit 10004 are connected in a state capable of transmitting data to each other. The processor 10001, memory 10002, and communication circuit 10004 are connected via an input / output interface 10003 so as to be capable of transmitting data to and from other hardware.

[0109] Alternatively, in the cloud cover forecasting devices 10, 10A, 10B, 10C and the power generation forecasting device 40, the learning units 11, 11A, 11B, 11C, the first cloud cover rate deriving unit 102 (102B, 102C), the first time series data generating unit 103 (103B, 103C), the second cloud cover rate deriving unit 202 (202B, 202C), the second time series data generating unit 203 (203B, 203C), the correlation calculation unit 301 (301B, 301C), the time The functions of the difference map generation unit 302 (302B, 302C), the prediction unit 12, 12A, 12B, 12C, the cloud movement direction derivation unit 401 (401B, 401C), the point selection unit 402 (402B, 402C), the third cloud coverage ratio derivation unit 403 (403B, 403C), the power generation amount prediction value acquisition unit 41, the learning model unit 42, and the control unit (not shown) may be realized by a dedicated processing circuit 20001, as shown in Figure 14.

[0110] The processing circuit 20001 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), an FPGA (Field-Programmable Gate Array), a SoC (System-on-a-Chip), or a system LSI (Large-Scale Integration). Alternatively, it may be a combination of these. Furthermore, the memory 20002 or another memory (not shown) implements the camera image storage unit 101 (101B, 101C), the time difference map storage unit 303 (303B, 303C), and a storage unit (not shown). The memory 20002 may be a non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable Read Only Memory), or flash memory, or a magnetic disk such as a hard disk or flexible disk, or an optical disk such as a CD (Compact Disc) or DVD (Digital Versatile Disc), or a magneto-optical disk. The communication circuit 20004 implements a communication unit (not shown). The processing circuit 20001 and the memory 20002 or the communication circuit 20004 are connected in a state where they can transmit data to each other. In addition, the processing circuit 20001, the memory 20002, and the communication circuit 20004 are connected in a state where they can transmit data to other hardware via the input / output interface 20003.In addition, in the cloud cover forecasting devices 10, 10A, 10B, 10C and the power generation amount forecasting device 40, the learning units 11, 11A, 11B, 11C, the first cloud cover ratio derivation units 102 (102B, 102C), the first time series data generation units 103 (103B, 103C), the second cloud cover ratio derivation units 202 (202B, 202C), the second time series data generation units 203 (203B, 203C), and the correlation calculation units 301 (301B, 301C) The functions of the time difference map generation unit 302 (302B, 302C), the prediction unit 12, 12A, 12B, 12C, the cloud movement direction derivation unit 401 (401B, 401C), the point selection unit 402 (402B, 402C), the third cloud coverage rate derivation unit 403 (403B, 403C), the power generation amount prediction value acquisition unit 41, the learning model unit 42, and the control unit (not shown) may be realized by separate processing circuits, or may be realized together by a processing circuit.

[0111] Alternatively, in the cloud cover forecasting devices 10, 10A, 10B, 10C and the power generation forecasting device 40, the learning units 11, 11A, 11B, 11C, the first cloud cover ratio deriving unit 102 (102B, 102C), the first time series data generating unit 103 (103B, 103C), the second cloud cover ratio deriving unit 202 (202B, 202C), the second time series data generating unit 203 (203B, 203C), the correlation calculation unit 301 (301B, 301C), the time difference map generating unit 302 (302 Some of the functions of the prediction units 12, 12A, 12B, 12C, the cloud movement direction derivation unit 401 (401B, 401C), the point selection unit 402 (402B, 402C), the third cloud coverage ratio derivation unit 403 (403B, 403C), the power generation amount forecast value acquisition unit 41, the learning model unit 42, and a control unit (not shown) may be realized by the processor 10001 and the memory 10002, and the remaining functions may be realized by the processing circuit 20001. In this way, each of the functions of the above components can be realized by hardware, software, firmware, or a combination of these.

[0112] It should be noted that, within the scope of this disclosure, the embodiments may be freely combined, any component of each embodiment may be modified, or any component of each embodiment may be omitted.

[0113] The present disclosure makes it possible to derive cloud cover prediction results using a smaller configuration than conventional methods, and is therefore suitable for use in a system including a cloud cover prediction device that predicts cloud cover used to predict power generation, and a power generation prediction device that predicts power generation from the cloud cover.

[0114] 1 (1A, 1B, 1C) Power generation forecasting system, 20 Camera, 30 Satellite data distribution system, 40 Power generation forecasting device, 50 Camera i (camera (i)), 10, 10A, 10B, 10C Cloud cover forecasting device, 11, 11A, 11B, 11C Learning unit, 101 (101B, 101C) Camera image storage unit, 102 (102B, 102C) First cloud cover rate deriving unit, 103 (103B, 103C) First time series data generating unit, 202 (202B, 202C) Second cloud cover rate deriving unit, 203 (203B, 203C) Second time series data generating unit, 301 (301B, 301C) Correlation calculation unit, 302 (302B, 302C) Time difference map generation unit, 303 (303B, 303C) Time difference map storage unit, 12, 12A, 12B, 12C Prediction unit, 401 (401B, 401C) Cloud movement direction derivation unit, 402 (402B, 402C) Point selection unit, 403 (403B, 403C) Third cloud coverage rate derivation unit, 1000 Time difference map, 10001 Processor, 10002 Memory, 10003 Input / output interface, 10004 Communication circuit, 20001 Processing circuit, 20002 Memory, 20003 Input / output interface, 20004 Communication circuit.

Claims

1. A cloud cover forecasting device comprising: a learning unit that calculates a correlation between first time series data of cloud cover ratio derived based on sky images, which are images of the sky photographed from an observation point, and second time series data of cloud cover ratio at multiple points derived based on satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the time series data with high correlation as a time difference map; and a prediction unit that uses the cloud movement direction obtained using satellite images and a preset time difference to derive the cloud cover ratio of a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud cover ratio.

2. The cloud amount forecasting device according to claim 1, wherein the learning unit comprises: a correlation calculation unit that performs a correlation calculation between the first time series data, which is time series data of cloud cover ratio derived based on a sky image, which is an image of the sky photographed from an observation point, and the second time series data, which is time series data of cloud cover ratio derived for a plurality of points using satellite imagery; a time difference map generation unit that uses the result of the correlation calculation to determine the time difference between highly correlated time series data for each of the plurality of points included in the satellite imagery; and a time difference map storage unit that stores a time difference map that indicates the correspondence between each of the plurality of points and the time difference between highly correlated time series data; and the prediction unit comprises: a cloud movement direction derivation unit that derives the movement direction of clouds using current frames of satellite imagery; a point selection unit that selects a point using the cloud movement direction and a preset time difference by referring to the time difference map; and a cloud cover ratio derivation unit that acquires a current cloud image and derives the cloud cover ratio at the selected point using the current cloud image.

3. The cloud cover forecasting device according to claim 2, characterized in that the current cloud image is a satellite image output by a satellite data distribution system that distributes data from artificial satellites.

4. The cloud cover forecasting device according to claim 2, characterized in that the current cloud images are images taken by cameras installed at multiple locations different from the observation location.

5. A power generation forecasting device characterized by inputting the predicted cloud coverage rate output by the cloud cover forecasting device described in claim 1 or claim 2 into a learning model that predicts power generation using the cloud coverage rate as an input, and obtaining the predicted power generation rate output by the learning model.

6. A power generation prediction device comprising: a learning unit that calculates a correlation between first time series data of cloud obscuration ratio derived based on sky images, which are images of the sky taken from an observation point, and second time series data of cloud obscuration ratio at multiple points derived based on satellite images, and accumulates a time difference map that indicates the correspondence between each point among the multiple points and the time difference between the highly correlated time series data; a prediction unit that uses the cloud movement direction obtained using the satellite images and a time difference received in advance to derive the cloud obscuration ratio of a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud obscuration ratio; and a power generation prediction value acquisition unit that inputs the predicted value of the cloud obscuration ratio into a learning model that predicts power generation using the cloud obscuration ratio as an input, and acquires a predicted value of the power generation output by the learning model.

7. A cloud amount forecasting method using a cloud amount forecasting device, wherein a learning unit of the cloud amount forecasting device calculates a correlation between first time series data of cloud obscuration rate derived based on sky images, which are images of the sky photographed from an observation point, and second time series data of cloud obscuration rate at multiple points derived based on satellite images, and accumulates, for each of the multiple points, a correspondence relationship between the point and the time difference between the time series data with high correlation as a time difference map, and a prediction unit of the cloud amount forecasting device uses the cloud movement direction obtained using the satellite images and a preset time difference to derive the cloud obscuration rate of a selected point by referring to the time difference map accumulated by the learning unit, and outputs a predicted value of the cloud obscuration rate.

8. A method for predicting power generation capacity using a power generation capacity prediction device, wherein a learning unit of the power generation capacity prediction device correlates time series data of cloud obscuration rate derived based on sky images, which are images of the sky photographed from an observation point, with time series data of cloud obscuration rate at multiple points derived from satellite images, and accumulates, as a time difference map, a correspondence relationship of time differences that has a high correlation with the points in the satellite images; a prediction unit of the power generation capacity prediction device derives the cloud obscuration rate of a selected point by referring to the time difference map accumulated by the learning unit using the cloud movement direction obtained from the satellite images and a preset time difference, and outputs a predicted value of the cloud obscuration rate; and a power generation capacity prediction value acquisition unit of the power generation capacity prediction device inputs the predicted value of the cloud obscuration rate output by the prediction unit into a learning model that predicts power generation using the cloud obscuration rate as an input, and acquires a predicted power generation value that is the prediction result output by the learning model.

Images