Device and method for predicting solar power generation amount

Abstract

To provide a solar power generation amount prediction device capable of accurately predicting a solar power generation amount from weather forecast data by performing machine learning, for example, for each building or dwelling unit to which a solar power generation panel is attached.SOLUTION: A solar power generation amount prediction device includes: a machine learning unit that performs, for each of one or a plurality of solar power generation panels, machine learning for each time based on a solar radiation amount actual measurement value and a power generation amount actual measurement value for each time in a past predetermined period with respect to a prediction target day; and a power generation amount prediction unit that predicts a power generation amount on the prediction target day based on results of the machine learning by the machine learning unit.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a solar power generation amount prediction device and a solar power generation amount prediction method.

Background Art

[0002] In recent years, the power supply and demand has been tight and the power price has fluctuated greatly. In such a situation, in order to buy and sell to make a profit in the power trading where the power trading has become active due to power liberalization, it is necessary to predict the power demand and supply at home in advance. As an essential element for that, it is important to accurately predict the solar power generation amount.

[0003] In the conventional method, the predicted value of the solar radiation amount in the weather forecast used for prediction is the global horizontal irradiance. This is the value of how much solar radiation enters a pyranometer placed on a horizontal plane without any obstacles. For example, when the solar altitude is low, the solar radiation becomes weak.

[0004] In the power generation amount prediction device etc. described in Patent Document 1, the power generation amount is predicted taking into account the change in the position of the sun (the change in the weather), and the power generation amount is predicted for each region (see Patent Document 1). In the solar power generation amount prediction method etc. described in Patent Document 2, the solar power generation amount in the solar power generation facility is calculated based on the global horizontal irradiance at clear sky without a single cloud for each time (see Patent Document 2), and such an ideal state model is required.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] However, in the conventional technology, for example, when there are obstacles to a solar power generation panel constituting a solar power generation system, due to environmental changes at each time of the day or environmental changes for each season, it may not be possible to accurately predict the solar power generation amount.

[0007] The present disclosure has been made in consideration of such circumstances, and an object of the present disclosure is to provide a solar power generation amount prediction device and a solar power generation amount prediction method that can accurately predict the solar power generation amount from weather forecast data by performing machine learning for each building or household equipped with a solar power generation panel, for example.

Means for Solving the Problems

[0008] One aspect of the present disclosure is a solar power generation amount prediction device including: a machine learning unit that performs machine learning for each time based on measured solar radiation amounts and measured power generation amounts for each time in a past predetermined period with respect to a prediction target day for each one or more solar power generation panels; and a power generation amount prediction unit that predicts the power generation amount on the prediction target day based on the result of the machine learning by the machine learning unit.

[0009] One aspect of the present disclosure is a solar power generation amount prediction method in which the machine learning unit performs machine learning for each time based on measured solar radiation amounts and measured power generation amounts for each time in a past predetermined period with respect to a prediction target day for each one or more solar power generation panel units or for each building or household, and the power generation amount prediction unit predicts the power generation amount on the prediction target day based on the result of the machine learning by the machine learning unit.

Advantages of the Invention

[0010] According to the photovoltaic power generation amount prediction device and the photovoltaic power generation amount prediction method according to the present disclosure, for example, for each building or household equipped with a photovoltaic panel, by performing machine learning, the photovoltaic power generation amount can be accurately predicted from weather forecast data.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0013] [Photovoltaic Power Generation Amount Prediction Device] FIG. 1 is a diagram showing an example of the configuration of the photovoltaic power generation amount prediction device 1 according to the embodiment. Note that detailed description of the photovoltaic power generation system (e.g., photovoltaic power generation panels, etc.) for which the photovoltaic power generation amount prediction device 1 predicts the power generation amount will be omitted. The photovoltaic power generation system is not particularly limited, and may be installed at any location, for example, in a privately-owned house, a company building, a building or its dwelling unit in an apartment building, or a public building such as a school.

[0014] Also, the photovoltaic power generation amount prediction device 1 may be owned by any person. For example, for each photovoltaic power generation system, it may be owned by the owner of the photovoltaic power generation system, or may be used as a common device (e.g., a server device may be sufficient) for a plurality of photovoltaic power generation systems.

[0015] In this embodiment, the photovoltaic power generation amount prediction device 1 is configured using, for example, a computer. The photovoltaic power generation amount prediction device 1 includes an input unit 11, an output unit 12, a storage unit 13, a calculation unit 14, and a facility control unit 15. The calculation unit 14 includes a machine learning unit 131 and a power generation amount prediction unit 132.

[0016] The input unit 11 inputs information. The input unit 11 has a function of inputting, for example, information output from an external device. In this embodiment, the input unit 11 inputs, for example, information used in machine learning of the power generation amount by photovoltaic power generation, and information used when predicting the power generation amount based on the machine learning result (machine learning model). The external device is not particularly limited, and may be, for example, an external sensor, an external database, or an external computer. Note that the input unit 11 may have an operation unit that accepts manual operations by the user.

[0017] The output unit 12 outputs information. The output unit 12 has a function of outputting information to an external device, for example. In this embodiment, the output unit 12 outputs, for example, information on the prediction result of the power generation amount. The external device is not particularly limited, and may be, for example, a display device that displays and outputs information on a screen. Note that the external device may be an external database or an external computer, etc.

[0018] Here, the input unit 11 may have a function of receiving information transmitted from an external device, for example. Also, the output unit 12 may have a function of transmitting information to an external device, for example. That is, the input unit 11 and the output unit 12 may have a function of communicating with an external device by wire or wirelessly.

[0019] The storage unit 13 stores information. In the present embodiment, the storage unit 13 stores the learning model group MG.

[0020] The machine learning unit 131 performs machine learning on the power generation amount by solar power generation. The power generation amount prediction unit 132 predicts the power generation amount based on the machine learning result (machine learning model).

[0021] Here, in the solar power generation amount prediction device 1 according to the present embodiment, the machine learning unit 131 performs machine learning for each time for a past predetermined period (learning period) with respect to the prediction target day (the day to be predicted) for each one or more solar power generation panels based on the measured solar radiation amount value and the measured power generation amount value for each time. The power generation amount prediction unit 132 predicts the power generation amount on the prediction target day based on the result of the machine learning by the machine learning unit 131. Therefore, in the solar power generation amount prediction device 1 according to the present embodiment, by performing machine learning, the solar power generation amount can be accurately predicted from the weather forecast data. Note that the solar power generation panel may be, for example, a stand-alone type (for example, one installed on the ground).

[0022] In addition, in the photovoltaic power generation amount prediction device 1 according to the present embodiment, for example, one or a plurality of photovoltaic panels constitute one or a plurality of photovoltaic panel units and are attached to a building or a dwelling unit. And the photovoltaic panel is connected to the electrical circuit inside the building or the dwelling unit. The machine learning by the machine learning unit 131 and the prediction by the power generation amount prediction unit 132 are performed for each building or each dwelling unit. Therefore, in the photovoltaic power generation amount prediction device 1 according to the present embodiment, by performing machine learning for each building or dwelling unit to which a photovoltaic panel is attached (that is, taking the building or dwelling unit as a unit), the photovoltaic power generation amount can be accurately predicted from the weather forecast data. Note that the electrical circuit may be various electrical circuits.

[0023] The facility control unit 15 has a function of controlling a predetermined facility based on the prediction result of the power generation amount by the power generation amount prediction unit 132. Note that the photovoltaic power generation amount prediction device 1 does not necessarily have to include the facility control unit 15.

[0024] Here, the photovoltaic power generation amount prediction device 1 includes, for example, a control unit that performs various processes and controls in the photovoltaic power generation amount prediction device 1. The control unit has, for example, a processor such as a CPU (Central Processing Unit), and performs various processes and controls by executing a predetermined program (computer program). Here, the program may be stored in the storage unit 13, for example. Note that the function of the facility control unit 15 may be regarded as a part of the function of the control unit, for example.

[0025] <Learning model group> FIG. 2 is a diagram showing an example of the configuration of the learning model group MG according to the embodiment. The learning model group MG includes a plurality of machine learning models. In the present embodiment, the learning model group MG includes 24 machine learning models M0 to M23. In this embodiment, one day (24 hours) is divided into 24 time slots, and a total of 24 machine learning models M0 to M23 are used corresponding to each time slot.

[0026] <Specific Example of Machine Learning> FIG. 3 is a diagram schematically showing a specific example of machine learning according to the embodiment. Let the number of learning days be L (L is an integer of 2 or more). In this example, L is set to about 20.

[0027] FIG. 3 shows a graph of the measured solar irradiance and a graph of the measured power generation. In the graph of the measured solar irradiance, the horizontal axis represents the passage of time, and the vertical axis represents the measured solar irradiance. The horizontal axis shows the Nth day, the (N + 1)th day, and the (N + 2)th day as the number of learning days. Note that the number of learning days is from the 1st day to the Lth day. In this example, N is an integer of 1 or more and (L - 2) or less. On each day, there are 24 hours from 0:00 to 24:00 (= 0:00). In the graph of the measured solar irradiance, the characteristics A(N) of the measured solar irradiance on the Nth day, the characteristics A(N + 1) of the measured solar irradiance on the (N + 1)th day, and the characteristics A(N + 2) of the measured solar irradiance on the (N + 2)th day are shown. Note that the same applies to other learning days, and the illustration is omitted. The characteristics A(1) to A(L) are the measured values (measurement values) for each corresponding day.

[0028] In the graph of the measured power generation, the horizontal axis represents the passage of time, and the vertical axis represents the measured solar irradiance. Here, the horizontal axis is the same as the horizontal axis of the graph of the measured solar irradiance. In the graph of the measured power generation, the characteristics B(N) of the measured power generation on the Nth day, the characteristics B(N + 1) of the measured power generation on the (N + 1)th day, and the characteristics B(N + 2) of the measured power generation on the (N + 2)th day are shown. Note that the same applies to other learning days, and the illustration is omitted. The characteristics B(1) to B(L) are respectively the measured values (measurement values) for the corresponding day.

[0029] In the example of FIG. 3, as machine learning for each time, machine learning for each time at 1-hour intervals is performed. Specifically, machine learning D(0) for the data at 0:00 to machine learning D(23) for the data at 23:00 are performed. In each of the machine learning D(0) for the data at 0:00 to machine learning D(23) for the data at 23:00, machine learning is performed using the set (each of the data groups C(0) to C(23)) of the information (in this example, the measured solar radiation value and the measured power generation value) at the corresponding time for each learning day (the 1st day to the Lth day) in the learning period (for L days).

[0030] In each of the machine learning D(0) for the data at 0:00 to machine learning D(23) for the data at 23:00, as a result of the machine learning, each learned model of the machine learning models M0 to M23 is obtained. Here, in each of the learned models of the machine learning models M0 to M23, a model is generated with the input as the solar radiation amount and the output as the power generation amount.

[0031] As a specific example, the machine learning D(12) for the data at 12:00 will be described. In the photovoltaic power generation prediction device 1, the machine learning unit 131 performs the process of the machine learning D(12) for the data at 12:00 using the data group C(12). The data group C(12) includes the data of the measured solar radiation value and the measured power generation value at 12:00 (12:00) for each day from the 1st day to the Lth day. The data group C(12) is used, for example, as teacher data for machine learning. The machine learning unit 131 generates a learned model of the machine learning model M12 corresponding to 12:00 by performing the process of the machine learning D(12) for the data at 12:00. Note that various methods may be used as the method for performing the machine learning of the machine learning models M0 to M23.

[0032] Here, the case where machine learning D(12) of the 12 o'clock data is performed using the data at 12 o'clock during the learning period has been described. Similarly, machine learning D(11) of the 11 o'clock data is performed using the data at 11 o'clock during the learning period, machine learning D(13) of the 13 o'clock data is performed using the data at 13 o'clock during the learning period, and the same applies to other times.

[0033] <Prediction of Power Generation Amount Based on Machine Learning Model> In the solar power generation amount prediction device 1, the power generation amount prediction unit 132 can calculate the predicted value of the power generation amount by solar power generation for each time using each of the machine learning models M0 to M23 that have been machine-learned. In this example, in the process of predicting the power generation amount, in each of the machine learning models M0 to M23, the predicted value of the solar irradiance (for example, the forecast value) is input, and the output is the predicted value of the power generation amount.

[0034] The power generation amount prediction unit 132 can perform the power generation amount prediction E1 for one day corresponding to the 24-hour period by predicting the power generation amount for 24 hours. The aggregated result for one day may be, for example, a combination of 24 machine learning results for 24 hours. In this case, it is considered that it is easy to grasp the peak and the like in one day. Here, in this example, the machine learning models M0 to M23 exist for each hour. Therefore, for the time between two adjacent times (for example, the time between 11 o'clock and 12 o'clock, or the time between 12 o'clock and 13 o'clock, etc.), the power generation amount prediction unit 132 may use the interpolation value (for example, the linear interpolation value) of the predicted values of these two times, or may use the predicted value of one of these two times (in this example, it is considered that the value continues), or may use other values.

[0035] <Solar Irradiance Used for Machine Learning and Power Generation Amount Prediction> In this embodiment, the solar irradiance used for machine learning and power generation amount prediction is not particularly limited. For example, the global horizontal irradiance may be used, or other solar irradiances may be used.

[0036] <Example of the progress of the process> FIG. 4 is a diagram schematically showing an example of the progress of the process according to the embodiment. In the example of FIG. 4, the horizontal axis represents the passage of time for one year (January to December), and the vertical axis schematically represents the progress of the process.

[0037] In this example, for each day when predicting the power generation amount, a learning period of about one month before that day is set. That is, in this example, the machine learning unit 131 performs machine learning processing of the machine learning models M0 to M23 for each day when predicting the power generation amount, and obtains the machine learning results of the machine learning models M0 to M23. And in this example, the power generation prediction unit 132 predicts the power generation amount for each day when predicting the power generation amount, using the machine learning results of the machine learning models M0 to M23 corresponding to that day.

[0038] In the example of FIG. 4, the machine learning processing progresses day by day. FIG. 4 schematically shows the progress of the process P1, and for some days (8 days in the example of FIG. 4), examples of the periods of the learning processes F1 to F8 and the prediction processes G1 to G8 are shown. As a specific example, in order to perform the prediction process G1 on a certain day, the learning process F1 of machine learning in a learning period of about one month before that day is performed. The same applies to other days.

[0039] In this way, in the present embodiment, for predicting the daily power generation amount for 365 days in a year (or 366 days in a leap year), machine learning processing for a predetermined period before each day is performed. Also, in the present embodiment, one day is divided into a predetermined number (24 in this example) of time points, and machine learning processing is performed for each time point. Accordingly, in this embodiment, for example, the same processing module can handle different seasons. That is, for example, although the environment (solar power generation environment) such as the solar altitude or solar azimuth changes depending on the season, the learning data (machine learning models M0 to M23) can always be updated with the latest data to cope with the environmental changes.

[0040] <Example of learning period> FIG. 5 is a diagram showing an example of the relationship between the number of training days and the coefficient of determination according to the embodiment. In the graph shown in FIG. 5, the horizontal axis represents the number of training days (learning period), and the vertical axis represents the coefficient of determination of the machine learning model. Characteristic 2011 representing an example of the relationship between the number of training days and the coefficient of determination is shown in the graph. In the example of FIG. 5, according to characteristic 2011, the number of training days of about 22 or 23 days is the best. In the example of FIG. 5, according to characteristic 2011, the number of training days from 15 to 25 days (or from 15 to 28 days, etc.) is good.

[0041] Here, in the example of FIG. 5, when the data to be learned is too small, the prediction accuracy tends to deteriorate, and when the learning data is increased, the prediction accuracy tends to improve. However, if the learning period is extended too much, the accuracy tends to deteriorate because seasonal changes (environmental changes) are included. Note that the example of FIG. 5 is for illustrative purposes and is not necessarily limited to this example. An appropriate learning period may be set during actual operation.

[0042] Thus, in the solar power generation amount prediction device 1 according to this embodiment, the predetermined period (learning period) may be 15 days or more and 25 days or less. Therefore, in the solar power generation amount prediction device 1 according to this embodiment, by setting an appropriate learning period, for example, it is possible to improve the accuracy of power generation amount prediction.

[0043] [Examples of obstacles to solar radiation] Referring to FIGS. 6 to 8, examples of obstacles to solar insolation are shown. FIG. 6 is a diagram showing an example of measurement of the solar insolation amount of the sun Q1 by the pyranometer 1011. In the example of FIG. 6, the case of measuring the total horizontal solar insolation amount by the pyranometer 1011 is shown.

[0044] FIG. 7 is a diagram showing an example of the relationship between the height of the sun and an obstacle. In the example of FIG. 7, solar power generation panels 1231 and 1232 are installed on the upper part (for example, the roof or the rooftop) of the house 1211. Note that the installation location of the solar power generation panel is not particularly limited, and as another example, it may be a veranda, the roof of a carport, window glass, or an outer wall surface, etc. Next to the house 1211, a taller house (the neighboring house 1221) exists as an obstacle. FIG. 7 shows the sun Q11 in a high altitude state and the sun Q12 in a lower altitude state than that. In the sun Q11 in a high altitude state, the solar insolation is not blocked by the neighboring house 1221, the solar insolation area R11 is wide, and the solar insolation hits most of the solar power generation panel 1231 and the entire solar power generation panel 1232. On the other hand, in the sun Q12 in a lower altitude state than that, the solar insolation is blocked by the neighboring house 1221, the solar insolation area R12 is narrow, the solar insolation hits about half of the solar power generation panel 1231, and hardly any solar insolation hits the solar power generation panel 1232.

[0045] Thus, for example, if there is a large neighboring house or the like on the south side or the like with respect to the solar power generation panel, the solar insolation is easily blocked. Also, when the solar altitude is low, the solar insolation is easily blocked. The solar altitude changes with time. As a specific example, if there is a high building on the south side or the like with respect to the solar power generation panel, the solar insolation with a low solar altitude in the morning and evening is easily blocked. In particular, in the season close to the winter solstice, the solar insolation with a low solar altitude in the morning and evening is easily blocked.

[0046] FIG. 8 is a diagram showing an example of the relationship between the azimuth of the sun and an obstacle. In the example of FIG. 8, a solar power generation panel 1411 is shown. FIG. 8 shows the sun Q21 and the solar radiation area R21 in the morning, the sun Q22 and the solar radiation area R22 at noon, and the sun Q23 and the solar radiation area R23 in the evening. In the example of FIG. 8, a large tree 1421 exists as an obstacle on the southwestern side of the solar power generation panel 1411. Due to this tree 1421, the solar radiation of the sun Q23 in the evening (a part of the solar radiation in the example of FIG. 8) is blocked, and the solar radiation hitting the solar power generation panel 1411 is reduced. Thus, for example, if there are high obstacles to the southwestern or southeastern side of a solar power generation panel, the solar radiation is likely to be blocked.

[0047] Here, as shown in the examples of FIGS. 7 and 8, even if the amount of solar radiation (measured value or forecast value) is the same, the amount of solar radiation hitting the solar power generation panel may vary depending on the surrounding situation for each moment of the day. Also, the amount of solar radiation hitting the solar power generation panel may vary, for example, for each season.

[0048] Here, generally, the longer the period of the learning data, the larger the number of data and the higher the prediction accuracy. However, if the period of the learning data is too long, since the season changes, the solar altitude and the solar azimuth shift even at the same time. From these facts, there is an appropriate period for the learning data, and as an example, it is considered appropriate that the period is about 15 to 25 days. Also, the environment around the solar power generation panel may change. For example, broad-leaved trees grow leaves and become lush in spring, generating shade, but shed their leaves in autumn to allow sunlight to pass through. Also, for example, if a high building is built next to the solar power generation panel, the solar radiation may be blocked and a shadow may be formed. Such environmental changes can cause a large change in the power generation amount by solar power generation. In order to cope with such environmental changes, in the prior art, it was necessary to change the prediction model one by one in the prediction calculation that modeled the surrounding situation. In contrast, in the predictive calculation using machine learning as in this embodiment, such environmental changes can be learned and followed, and it is not always necessary to change the prediction model one by one. In this embodiment, by setting an appropriate learning period, the response to such environmental changes can be realized with higher accuracy.

[0049] [Modification Example: Use Other Than Solar Irradiance Amount] Here, in this embodiment, as an example where a machine learning model M0 to M23 that outputs the power generation amount with the solar irradiance amount as the input is used, as another example, other information may be used as the input together with the solar irradiance amount. The other information may be, for example, one or both of temperature information and cloud amount information. As a specific example, during machine learning, in the machine learning model, the measured value of the solar irradiance amount is used as the input, and one or both of the measured value of the temperature and the measured value of the cloud amount are used. And at the time of predicting the power generation amount, in the machine learning model, the predicted value (for example, the forecast value) of the solar irradiance amount is used as the input, and one or both of the predicted value (for example, the forecast value) of the temperature and the predicted value (for example, the forecast value) of the cloud amount are used. Thereby, a predicted value of the power generation amount by solar power generation is output from the machine learning model.

[0050] In this way, in the solar power generation amount prediction device 1 according to this embodiment, the machine learning unit 131 may further perform machine learning based on at least one of the measured temperature value and the measured cloud amount value. Therefore, in the solar power generation amount prediction device 1 according to this embodiment, by using other information together with the solar irradiance amount, for example, it is possible to improve the accuracy of power generation amount prediction.

[0051] [Multiple Times in One Day] Here, in this embodiment, as an example where the times at one-hour intervals are used as the multiple times in one day, the interval between adjacent times is not necessarily limited to this example. For example, times at an arbitrary interval such as times at 30-minute intervals or times at two-hour intervals may be used. In addition, in this embodiment, a case where well-separated times such as 0:00 to 23:00 are used has been shown, but the present invention is not limited to this, and any time may be used. In addition, in this embodiment, a case where the interval between two adjacent times is constant (1 hour in this embodiment) has been shown. As another example, a plurality of times where the interval between two adjacent times is not constant, such as 10:00, 11:30, 12:30, and 13:00, may be used. Generally, as such an interval (time interval) becomes shorter, the accuracy increases, but the calculation time becomes longer. Conversely, as such an interval (time interval) becomes longer, the accuracy decreases, but the calculation time becomes shorter. Therefore, such an interval may be determined based on conditions such as the performance of a computer (the solar power generation amount prediction device 1 in this embodiment).

[0052] As described above, in the solar power generation amount prediction device 1 according to this embodiment, for example, for each building or household equipped with a solar power generation panel, by performing machine learning, the solar power generation amount can be accurately predicted from weather forecast data (for example, the forecast value of solar radiation amount, etc.).

[0053] In the solar power generation amount prediction device 1 according to this embodiment, for example, in order to consider the influence of obstacles, the solar radiation amount forecast value at a certain time in the weather forecast is not used for predicting the power generation amount at all times. Instead, it is classified for each time, and the weather forecast value at the same time is used for predicting the solar power generation amount at each time. Thereby, in the solar power generation amount prediction device 1 according to this embodiment, it is possible to accurately predict the solar power generation amount in consideration of the influence of the surrounding environment of the solar power generation panel.

[0054] [Example of a power generation amount management system constructed for each building] FIG. 9 is a diagram showing an example of a power generation amount management system constructed for each building. In this example, a power generation amount management system constructed for each building is illustrated, but as another example, it may be applied to a power generation amount management system constructed for each household. In the example of FIG. 9, a solar power generation panel unit 331, a distribution board 332, equipment 333, a storage unit 411, a display unit 412, and a solar power generation amount prediction device 1a are installed in a building 311. A measuring instrument 351 is attached to the distribution board 332.

[0055] Here, the solar power generation amount prediction device 1a generally has the same components as the solar power generation amount prediction device 1 shown in FIG. 1. Therefore, in the present embodiment, for convenience of explanation, the same reference numerals as those of the components shown in FIG. 1 are used for explanation.

[0056] The building 311 may be any building, for example, a private house (which may be called a mansion or a residence), a company building, an apartment building, or a public building such as a school. The solar power generation panel unit 331 collectively represents the solar power generation panels provided in the building 311. For example, it may have only one solar power generation panel, or may have a plurality of solar power generation panels.

[0057] The distribution board 332 is the distribution board of the solar power generation panel unit 331. Note that the distribution board 332 may have both the function of the distribution board of the solar power generation panel unit 331 and the function of other distribution boards. The measuring instrument 351 measures the power generation amount of the solar power generation panel unit 331 as power data.

[0058] The equipment 333 may be any equipment, for example, a storage battery, air conditioning equipment, or other household appliances. The storage battery may be, for example, a storage battery that stores the power generated by the solar power generation panel unit 331. In this example, the equipment 333 can be controlled from the outside and is the equipment to be controlled. In this example, for simplicity of explanation, one piece of equipment 333 is shown, but two or more pieces of equipment may be the control targets. Note that the facility 333 may be arranged, for example, inside the building 311 or may be arranged as an attachment outside the building 311.

[0059] The storage unit 411 stores (memorizes) the power data measured by the measuring instrument 351. The display unit 412 is, for example, a display device having a screen, and displays the information output by the output unit 12 of the solar power generation amount prediction device 1a on the screen. In the example of FIG. 9, the storage unit 411 and the display unit 412 are shown as components separate from the solar power generation amount prediction device 1a. However, as another example, one or both of the functions of the storage unit 411 and the display unit 412 may be included as functions of the solar power generation amount prediction device 1a.

[0060] In the solar power generation amount prediction device 1a, the power data stored in the storage unit 411 is input by the input unit 11, and the power generation amount is predicted by the power generation amount prediction unit 132 based on the power data. Further, in the solar power generation amount prediction device 1a, the information on the prediction result of the power generation amount is output by the output unit 12 to the display unit 412, and thereby the prediction result is displayed by the display unit 412. Further, in the solar power generation amount prediction device 1a, based on the predicted power generation amount, the facility control unit 15 controls the facility 333 at a predetermined timing. The predetermined timing is not particularly limited, and may be, for example, a constant interval in time or may be another interval. The predetermined timing may be, for example, an interval regarded as real time, or may be a sequential interval that is not real time.

[0061] Here, the timing at which the power data is sent from the measuring instrument 351 to the storage unit 411 is not particularly limited, and various timings may be used. For example, the power data collected at a predetermined granularity such as every 10 minutes or every 30 minutes from the measuring instrument 351 may be used for storage. In the example of FIG. 9, the storage unit 411, the photovoltaic power generation amount prediction device 1a, and the display unit 412 are provided in the building 311 and are used like an edge terminal.

[0062] As described above, in the power generation amount management system according to this example, in the photovoltaic power generation amount prediction device 1a installed in the building 311, the machine learning unit 131 uses the power data stored in the storage unit 411 that stores the power data collected at a predetermined granularity from the measuring instrument 351 installed in the building 311 to perform machine learning. Further, the photovoltaic power generation amount prediction device 1a displays the prediction result by the power generation amount prediction unit 132 on the display unit 412. Therefore, in the photovoltaic power generation amount prediction device 1a, machine learning can be performed using the power data from the measuring instrument 351 installed in the building 311, and the prediction result of the power generation amount can be notified to users and the like. Here, in this example, the case where the photovoltaic power generation amount prediction device 1a and the measuring instrument 351 are installed in the building 311 is shown. However, as another example, they may be installed in a residential unit instead of the building 311.

[0063] Also, in the power generation amount management system according to this example, in the photovoltaic power generation amount prediction device 1a, the facility control unit 15 controls the facility 333 installed in the building 311 at a predetermined timing based on the prediction result by the power generation amount prediction unit 132. Therefore, in the photovoltaic power generation amount prediction device 1a, the facility 333 of the building 311 can be appropriately controlled based on the prediction result of the power generation amount. Here, in this example, the case where the facility 333 is installed in the building 311 is shown. However, as another example, it may be installed in a residential unit instead of the building 311.

[0064] [Example of a power generation amount management system constructed using the cloud] FIG. 10 is a diagram showing an example of a power generation amount management system constructed using the cloud. Here, in this example, for the convenience of explanation, the same components as those in the example of FIG. 9 are denoted by the same reference numerals as those in the example of FIG. 9 and are described.

[0065] In the example of FIG. 10, a solar power generation panel unit 331, a distribution board 332a, equipment 333, and a display unit 412 are installed in a building 311. A measuring instrument 351 is attached to the distribution board 332a. In addition, the distribution board 332a is provided with a communication unit 511. As another example, the communication unit 511 may be provided separately from the distribution board 332a.

[0066] In this example, a solar power generation amount prediction device 1b, a storage unit 631, and a display unit 632 are provided in a cloud region (cloud unit W1). The communication unit 511 in the building 311 can communicate with the cloud unit W1 via a communication line V1. This communication may be, for example, wired communication or wireless communication.

[0067] The communication unit 511 transmits the power data measured by the measuring instrument 351 to the cloud unit W1. The storage unit 631 stores (memorizes) the power data transmitted from the communication unit 511. The display unit 632 is a display device having a screen, for example, and displays the information output by the output unit 12 of the solar power generation amount prediction device 1b on the screen. In the example of FIG. 10, the storage unit 631 and the display unit 632 are shown as components separate from the solar power generation amount prediction device 1b. However, as another example, one or both of the functions of the storage unit 631 and the display unit 632 may be included as functions of the solar power generation amount prediction device 1b.

[0068] In the solar power generation amount prediction device 1b, the power data stored in the storage unit 631 is input by the input unit 11, and the power generation amount is predicted by the power generation amount prediction unit 132 based on the power data. Further, in the solar power generation amount prediction device 1b, the information on the prediction result of the power generation amount is output by the output unit 12 to the display unit 412 and the display unit 632, and thereby the prediction result is displayed by the display unit 412 and the display unit 632. Here, in this example, the information from the photovoltaic power generation amount prediction device 1b to the display unit 412 is transmitted via the communication line V1. Note that, in this example, a case is shown where the prediction results of the power generation amount are displayed by both the display unit 412 installed in the building 311 and the display unit 632 existing in the cloud unit W1. However, for example, either one of the display units may not be provided.

[0069] Also, in the photovoltaic power generation amount prediction device 1b, based on the predicted power generation amount, the facility control unit 15 controls the facility 333 at a predetermined timing. Here, in this example, the information from the photovoltaic power generation amount prediction device 1b to the facility 333 is transmitted via the communication line V1. The predetermined timing is not particularly limited. For example, it may be at regular time intervals, or at other intervals. The predetermined timing may be, for example, an interval regarded as real time, or a sequential interval that is not real time.

[0070] Here, the timing at which the power data is sent from the measuring instrument 351 to the communication unit 511 (or the storage unit 411) is not particularly limited, and various timings may be used. For example, the power data collected at a granularity such as every 10 minutes or every 30 minutes from the measuring instrument 351 may be used for communication and storage. Note that, in the example of FIG. 10, the storage unit 631, the photovoltaic power generation amount prediction device 1b, and the display unit 632 are provided in the cloud (cloud unit W1).

[0071] As described above, in the power generation amount management system according to this example, in the photovoltaic power generation amount prediction device 1b provided in the cloud, the machine learning unit 131 performs machine learning using the power data stored in the storage unit 631 that stores the power data collected at a predetermined granularity from the measuring instrument 351 installed in the building 311. Also, the photovoltaic power generation amount prediction device 1b displays the prediction results by the power generation amount prediction unit 132 on the display units 412 and 632. Therefore, the photovoltaic power generation prediction device 1b can perform machine learning using the power data from the measuring instrument 351 installed in the building 311, and can also notify the user or the like of the prediction result of the power generation amount. Here, in this example, the case where the measuring instrument 351 is installed in the building 311 is shown. However, as another example, it may be installed in a residential unit instead of the building 311.

[0072] Also, in the power generation amount management system according to this example, in the photovoltaic power generation prediction device 1b, the facility control unit 15 controls the facility 333 installed in the building 311 at a predetermined timing based on the prediction result by the power generation amount prediction unit 132. Therefore, the photovoltaic power generation prediction device 1b can appropriately control the facility 333 of the building 311 based on the prediction result of the power generation amount. Here, in this example, the case where the facility 333 is installed in the building 311 is shown. However, as another example, it may be installed in a residential unit instead of the building 311.

[0073] In addition, a program for realizing the functions of any component in any of the devices described above may be recorded on a computer-readable recording medium, and the program may be read into a computer system and executed. Here, the "computer system" is assumed to include an operating system or hardware such as peripheral devices. Also, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD (Compact Disc)-ROM (Read Only Memory), or a storage device such as a hard disk built into a computer system. Furthermore, the "computer-readable recording medium" also includes a volatile memory inside a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, and holds the program for a certain period of time. The volatile memory may be, for example, a RAM (Random Access Memory). The recording medium may be, for example, a non-transitory recording medium.

[0074] Also, the above program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network such as the Internet or a communication line such as a telephone line. Also, the above program may be for realizing a part of the functions described above. Furthermore, the above program may be a so-called difference file that can be realized in combination with a program already recorded in a computer system for the functions described above. The difference file may be called a difference program.

[0075] Also, the functions of any component in any of the devices described above may be realized by a processor. For example, each process in the embodiment may be realized by a processor that operates based on information such as a program and a computer-readable recording medium that stores information such as a program. Here, the processor may be realized by, for example, individual hardware for each function, or the functions of each part may be realized by integrated hardware. For example, the processor includes hardware, and the hardware may include at least one of a circuit that processes digital signals and a circuit that processes analog signals. For example, the processor may be configured using one or more circuit devices mounted on a circuit board, or one or both of one or more circuit elements. As the circuit device, an IC (Integrated Circuit) or the like may be used, and as the circuit element, a resistor or a capacitor or the like may be used.

[0076] Here, the processor may be, for example, a CPU. However, the processor is not limited to the CPU, and various processors such as a GPU (Graphics Processing Unit) or a DSP (Digital Signal Processor) may be used. Also, the processor may be, for example, a hardware circuit by an ASIC (Application Specific Integrated Circuit). Also, the processor may be configured by, for example, a plurality of CPUs, or may be configured by a hardware circuit by a plurality of ASICs. Also, the processor may be configured by, for example, a combination of a plurality of CPUs and a hardware circuit by a plurality of ASICs. Also, the processor may include, for example, one or more of an amplifier circuit or a filter circuit that processes analog signals.

[0077] As described above, the embodiments of this disclosure have been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and designs and the like within the scope not departing from the gist of this disclosure are also included.

[0078] [Appendix] (Configuration Example 1) to (Configuration Example 7) are shown.

[0079] (Configuration Example 1) For each one or more solar power generation panels, a machine learning unit that performs machine learning for each time based on the measured solar radiation amount and the measured power generation amount for each time in a past predetermined period with respect to the prediction target date, A power generation amount prediction unit that predicts the power generation amount on the prediction target date based on the result of the machine learning by the machine learning unit, A solar power generation amount prediction device comprising:

[0080] (Configuration Example 2) One or more of the solar power generation panels constitute one or more solar power generation panel units and are attached to a building or a dwelling, and are connected to an electrical circuit inside the building or the dwelling, The machine learning by the machine learning unit and the prediction by the power generation amount prediction unit are performed for each building or dwelling, The solar power generation amount prediction device according to (Configuration Example 1).

[0081] (Configuration Example 3) The machine learning unit performs the machine learning using the power data stored in an accumulation unit that accumulates the power data collected at a predetermined granularity from a measuring instrument installed in the building or the dwelling, The solar power generation amount prediction device displays the prediction result by the power generation amount prediction unit on a display unit. The solar power generation amount prediction device according to (Configuration Example 2).

[0082] (Configuration Example 4) The solar power generation amount prediction device further includes a facility control unit that controls facilities installed in the building or the dwelling at a predetermined timing based on the prediction result by the power generation amount prediction unit. The photovoltaic power generation amount prediction device described in (Configuration Example 2) or (Configuration Example 3).

[0083] (Configuration Example 5) The machine learning unit further performs the machine learning based on at least one of the measured air temperature value and the measured cloud amount value. The photovoltaic power generation amount prediction device according to any one of (Configuration Example 1) to (Configuration Example 4).

[0084] (Configuration Example 6) The predetermined period is 15 days or more and 25 days or less. The photovoltaic power generation amount prediction device according to any one of (Configuration Example 1) to (Configuration Example 5).

[0085] For example, it is also possible to provide a method performed in the photovoltaic power generation amount prediction device. (Configuration Example 7) The machine learning unit performs machine learning for each time based on the measured solar radiation amount value and the measured power generation amount value for each time in a past predetermined period with respect to the prediction target day, for each one or more photovoltaic panel units or for each building or dwelling unit. The power generation amount prediction unit predicts the power generation amount on the prediction target day based on the result of the machine learning by the machine learning unit. Photovoltaic power generation amount prediction method.

Explanation of reference numerals

[0086] 1, 1a, 1b... Photovoltaic power generation prediction device, 11... Input unit, 12... Output unit, 13... Memory unit, 14... Calculation unit, 15... Equipment control unit, 131... Machine learning unit, 132... Power generation prediction unit, 311... Building, 331... Photovoltaic panel unit, 332, 332a... Distribution board, 333... Equipment, 351... Measuring instrument, 411, 631... Storage unit, 412, 632... Display unit, 511... Communication unit, 1011... Pyranometer, 1211... House, 1221... Neighboring house, 1231~1232, 1411... Photovoltaic panel, 1421... Trees, 2011... Characteristics, A(1)~A(L)... Characteristics of measured solar irradiance values, B(1)~B(L)... Characteristics of measured power generation values, C(0)~C(23)... Data group, D(0)~D(23)... Machine learning of data for each time, E1... Prediction of daily power generation, F1~F8... Learning process, G1~G8... Prediction process, MG... Learning model group, M0~M23... Machine learning model, P1... Elapsed time of processing, Q1, Q11~Q12, Q21~Q23... Sun, R11~R12, R21~R23... Region, V1... Communication line, W1... Cloud part

Claims

1. For each one or more solar power generation panels, a machine learning unit that performs machine learning for each time based on the measured solar radiation amount and the measured power generation amount for each time in a past predetermined period with respect to the prediction target date; A power generation amount prediction unit that predicts the power generation amount on the prediction target date based on the result of the machine learning by the machine learning unit; A solar power generation amount prediction device comprising:

2. One or more of the solar power generation panels constitute one or more solar power generation panel units and are attached to a building or a dwelling unit, and are connected to an electric circuit inside the building or the dwelling unit. The machine learning by the machine learning unit and the prediction by the power generation amount prediction unit are performed for each building or each dwelling unit. The solar power generation amount prediction device according to claim 1.

3. The machine learning unit performs the machine learning using the power data stored in a storage unit that stores the power data collected at a predetermined granularity from a measuring instrument installed in the building or the dwelling unit. The solar power generation amount prediction device displays the prediction result by the power generation amount prediction unit on a display unit. The solar power generation amount prediction device according to claim 2.

4. The solar power generation amount prediction device further includes a facility control unit that controls facilities installed in the building or the dwelling unit at a predetermined timing based on the prediction result by the power generation amount prediction unit. The solar power generation amount prediction device according to claim 2 or claim 3.

5. The machine learning unit further performs the machine learning based on at least one of the measured air temperature value and the measured cloud amount value. The solar power generation amount prediction device according to any one of claims 1 to 3.

6. The predetermined period is 15 days or more and 25 days or less. The solar power generation amount prediction device according to any one of claims 1 to 3.

7. The machine learning unit performs machine learning for each time based on the measured solar radiation amount and the measured power generation amount for each time in a past predetermined period with respect to the prediction target date for each one or more solar power generation panel units or for each building or each dwelling unit. The power generation amount prediction unit predicts the power generation amount on the prediction target date based on the result of the machine learning by the machine learning unit. A solar power generation amount prediction method.

Images