JP5940922B2 - Renewable energy prediction device - Google Patents

Description

The present invention relates to a natural energy amount prediction apparatus that is applied to a natural energy prediction system including a power generation system and a weather data providing center and predicts an energy amount per unit time obtained from natural energy.

  2. Description of the Related Art Conventionally, for example, an information processing apparatus and its program as shown in Patent Document 1 below are known. This conventional information processing apparatus, etc., at least two days in the past that have similar weather transitions to the days for which the amount of power generation is predicted from the weather information indicating past weather transitions stored in the information storage unit. Similar date selection means to be selected, and time series data generation means for acquiring the power generation amount data for the day selected by the similar date selection means from the information holding unit, and generating the acquired power generation amount data as time series data for every fixed time And mathematical model generation means for generating a mathematical model that at least shows the correlation of the power generation amount between the times selected by the similar day selection means using time series data, and applying the mathematical model to a predetermined probability density function Power generation scenario that shows the predicted value of power generation per fixed time of the power generation facility by determining the probability distribution of power generation per fixed time and generating multiple sets of random numbers according to the probability distribution The adapted and a scenario generation unit that generates a plurality.

  Conventionally, for example, a power generation amount prediction device, a prediction method, and a prediction program as shown in Patent Document 2 below are also known. This conventional power generation amount predicting device, etc. includes a storage unit that stores past data related to the past power generation amount of the generator, and a statistical correlation between different times in the past data or a statistical correlation between different generator positions. And a predicted value calculation unit that calculates a predicted value related to the power generation amount of the generator as time series data including the occurrence probability.

  Furthermore, conventionally, for example, a power generation amount prediction system and a power generation amount prediction method of a solar power generation device as shown in Patent Document 3 below are also known. This conventional photovoltaic power generation amount prediction system calculates the cloud water particle size distribution by simulating a physical weather model related to the amount of water vapor and clouds in the atmosphere based on the weather forecast data, and based on the cloud water particle size distribution. Calculate the light transmittance data of the cloud at the predicted power generation amount, calculate the spectral solar radiation based on the cloud light transmittance data, and install the solar power generation installed based on the calculated spectral solar radiation The amount of power generation according to the device is predicted.

JP 2012-23816 A JP 2011-200040 A JP 2011-159199 A

  By the way, in the conventional information processing apparatus etc. shown in the said patent document 1, the electric power generation amount prediction apparatus shown in the said patent document 2, etc., the prediction value is obtained by selecting a plurality of past similar dates and taking their statistics. By calculating a probability distribution (for example, multi-dimensional normal distribution, beta distribution, etc.), or by taking a statistical correlation between different times in past data, the predicted value related to the power generation amount of the generator is calculated. It has become. However, for example, in the conventional information processing apparatus shown in Patent Document 1, if the similarity criterion is strict, many similar days cannot be selected, and if the criterion is relaxed, irrelevant days are also selected. It comes to be. Moreover, when there is no similar day, it becomes impossible to obtain the predicted value and its distribution.

  Also, for example, in the conventional information processing apparatus shown in Patent Document 1 described above, when a multidimensional normal distribution is used, the multidimensional normal distribution does not take into account the asymmetry of the prediction error. It is difficult to accurately calculate (predict deviation). With regard to this point, for example, in the conventional information processing apparatus shown in Patent Document 1, there is a description that the beta distribution can be used as the probability distribution, but the specific calculation contents and the like are disclosed. It has not been.

  The present invention has been made to cope with the above-described problems, and its purpose is to accurately calculate the predicted value of the energy amount per unit time obtained from natural energy, including the distribution of the prediction error. An object of the present invention is to provide an apparatus for predicting the amount of natural energy.

In order to achieve the above object, a natural energy amount prediction apparatus according to the present invention is a power generation system that converts an energy amount obtained from natural energy that changes due to a meteorological phenomenon into electric energy, supplies it to a power consuming device, and charges a storage battery. A power generation system that controls charging / discharging of the storage battery using prediction information of a natural energy amount per unit time obtained from the natural energy, and meteorological data related to a meteorological phenomenon affecting the natural energy are provided. The present invention is applied to a natural energy prediction system including a weather data providing center, and is communicably connected to the power generation system and the weather data providing center, and predicts the amount of natural energy using weather data from the weather data providing center. Predictive means is provided.

Features of the energy amount prediction apparatus according to the present invention, the are those predicted distribution of natural energy is represented by the asymmetric probability distribution, the prediction means, a meteorological data in areas where the power generation system is installed, Obtained as an explanatory variable related to the change in the amount of natural energy from the meteorological data providing center, and using the obtained explanatory variable, related to the first parameter corresponding to the expected value of the probability distribution and the spread of the probability distribution A second parameter is calculated, the probability distribution representing the predicted distribution of the natural energy amount is calculated using the calculated first parameter and the second parameter, and the natural energy amount is calculated based on the calculated probability distribution. Predicting the predicted value of the natural energy amount as the predicted information of the natural energy. It is to output to the stem.

In this case, for example, the power generation system predicts a power generation amount based on the prediction information of the natural energy amount, and controls charging / discharging of the storage battery according to the predicted power generation amount. In this case, the explanatory variables include, for example, temperature, atmospheric pressure, humidity, precipitation, precipitation, snowfall, snow cover, low cloud cover, middle cloud cover, high cloud cover, wind direction, wind speed, updraft, weather, and clear weather. It is possible to acquire and use at least one of the weather forecast values of the degree, the air mass and the solar radiation amount outside the atmosphere, and the lead time from the prediction execution time to the prediction target time. In this case, at least one of the weather forecast values, the lead time, the current situation, and the weather observation values observed in the past can be acquired and used. Further, in this case, an asymmetric probability distribution representing the predictive distribution of the natural energy, it is possible to Beta distribution or zero excess beta distribution. In these cases, the natural energy may be at least one natural energy of solar energy, solar thermal energy, wind energy, tidal energy, and water energy.

Also, in these cases, the predicting means calculates the mode value that appears most frequently in the calculated probability distribution, and follows the mode value calculated and the mode value in the probability distribution. At least the mode value among the frequently appearing values can be used as the predicted value of the natural energy amount .

Also, when these, the prediction means, which is a set of values appearing in a predetermined frequency by the probability distribution comprises a mode value most frequently occurring in the probability distribution that the calculated even without low period, the The confidence interval of the predicted value of the natural energy amount can be determined, and the determined confidence interval can be included in the prediction information of the natural energy amount and output to the power generation system .

In the case of these, the predicting means, based on the probability distributions calculated before Symbol meter, said a confidence interval outside the predicted values of the natural energy, the contrary to the predicted value of the natural energy, is measured were actual natural energy amount is less than the natural energy of a preset lower limit, or to calculate the probability of the actual natural energy is generated in the preset upper side of the natural energy or become mispredicted The probability of occurrence of the calculated misprediction can be included in the prediction information of the natural energy amount and output to the power generation system .

According to these, the predicting means acquires the weather data in the area where the power generation system is installed as an explanatory variable related to the change in the amount of natural energy from the weather data providing center, and uses the acquired explanatory variable, calculating a second parameter associated with the first parameter and spread of distribution corresponding to the expected value of the distribution, asymmetric probability distribution is calculated using these first parameter and the second parameter, specifically, A predicted distribution of the natural energy amount is represented by a beta distribution, a zero excess beta distribution, or the like, and a predicted value of the natural energy amount can be calculated based on this predicted distribution, that is, a probability distribution. As a result, parameters can be calculated using explanatory variables such as weather forecast values, so that the relationship between continuous variables can be described, for example, parameters of probability distribution by regression analysis that can be described by functions. The predicted distribution (occurrence distribution) can be calculated for any weather condition without selecting a similar date from the past data.

Further, by using the second parameter associated with the first parameter and spread of distribution corresponding to the expected value of the distribution, it is possible to calculate the asymmetric probability distribution, for example, per unit time obtained from solar energy Even when taking a value of a finite interval with a lower limit of “0” and an upper limit of the amount of solar radiation outside the atmosphere, like the amount of solar radiation as the amount of solar radiation, the tail of the distribution can be expressed accurately. It becomes possible. Therefore, a misprediction in which a predicted value greatly deviates can be calculated with extremely high accuracy and appropriately predicted.

Further, it is possible to the mode value in the probability distribution and the predicted value of the natural energy, when the subject is the prediction error in the asymmetric probability distribution can be represented accurately likelihood of occurrence of the event. In addition, since the confidence interval can be determined so as to include the mode value, it is possible to present the prediction very easily.

  Therefore, according to the natural energy amount prediction apparatus of the present invention, the predicted value of the natural energy amount including the distribution of the prediction error can be calculated with extremely high accuracy. Then, the probability of occurrence of a rare phenomenon in which the predicted value of the natural energy amount is erroneously predicted (the predicted value greatly deviates) can be predicted quantitatively with high accuracy.

Another feature of the natural energy amount prediction apparatus according to the present invention is that the prediction means acquires a third parameter representing a correlation between the prediction distributions of the natural energy amount at different time points corresponding to each of the different time points. Calculated using explanatory variables, and using the calculated third parameter, calculate a joint function that gives a correlation between the predicted distributions of the natural energy amount at the different time points, and according to the calculated third parameter, When a correlation is recognized between the predicted distributions of the natural energy amount at different time points, the calculated junction function is used to calculate the natural energy amount at the previous time point among the predicted distributions of the natural energy amount at the different time points. The natural energy that has been predicted at a later point in time based on the actual amount of natural energy measured corresponding to the predicted distribution There also be reflected in the predicted value.

According to these, the third parameter representing the correlation between the predicted distributions of the natural energy amount at different time points can be calculated using the explanatory variables at the different time points, and the natural energy at different time points can be calculated using the third parameter. A joining function can be calculated that gives a correlation between the predicted distributions of quantities. Then, according to the magnitude of the third parameter, that is, the strength of the correlation, for example, when a correlation between time points that are continuous in time is recognized, the correlation is measured corresponding to the predicted distribution of the natural energy amount at the previous time point. it is reflected in the predicted value of the natural energy, which is predicted at a later time the actual natural energy amount can and Turkey.

  Therefore, even in this case, the predicted value of the natural energy amount can be calculated with a very high accuracy including the distribution of the prediction error. Then, the probability of occurrence of a rare phenomenon in which the predicted value of the natural energy amount is erroneously predicted (the predicted value greatly deviates) can be predicted quantitatively with high accuracy.

Furthermore, in these cases, for example, when the amount of natural energy is the amount of solar radiation per unit time obtained from the solar energy, the predicting means includes the predicted distribution of the amount of solar radiation represented by the probability distribution and the using the predicted values of the solar radiation amount predicted based on the probability distribution, and predicted power generation amount distribution and power generation amount of the solar power generation system for converting the solar energy into electrical energy, the predicted power generation amount distribution及based on beauty power generation amount, it said photovoltaic systems, other power generation systems, also to coordinate the energy storing device and the power consuming device. According to these, it becomes possible to use electric energy converted (generated) from solar energy, which is natural energy, very efficiently.

It is the schematic which shows the structure of the natural energy amount prediction system which can apply the natural energy amount prediction apparatus which concerns on this invention. It is the schematic which concerns on embodiment of this invention and shows the structure of the solar energy power generation system of FIG. It is the schematic which shows the structure of the electric power generation monitor apparatus of FIG. It is the schematic which shows the structure of the management center of FIG. FIG. 5 is a functional block diagram functionally representing computer program processing executed by the control device (computer) of FIG. 4. It is a figure for demonstrating the parameter relevant to the spread of probability distribution. The asymmetric distribution according to the present embodiment is a diagram for describing the probability density distribution allowable. It is a figure for demonstrating that the probability density distribution which concerns on this embodiment has a mode value and an expected value distinguishable. It is a figure for demonstrating the example of a prediction when assuming that the distribution of predicted solar radiation amount has the prediction error according to the normal distribution which is a symmetrical distribution of an infinite interval. Relates to the present embodiment, the distribution of the predicted amount of solar radiation is a diagram for explaining a prediction example where the has a prediction error according Beta distribution is a distribution that tolerate the a and asymmetric finite interval. The relationship between the predicted solar radiation amount (predicted value) and the actually measured solar radiation amount (actual value) is shown, and (a) uses the expected value in the probability density distribution (probability distribution) as the predicted solar radiation amount (predicted value). (B) is a figure which shows the case where the mode value in probability density distribution (probability distribution) is used as a predicted solar radiation amount (predicted value). It is a diagram for explaining a confidence interval and erroneous predicted probability relative to the most frequent value of the asymmetric distribution according to the present embodiment in the probability density distribution for allowable. It is a diagram for explaining a confidence interval relative to the expected value and the median value in a probability distribution for allowable asymmetric distribution. When the confidence interval based on the expected value or median is determined for the predicted solar radiation (predicted value) and when the confidence interval based on the mode is determined for the predicted solar radiation (predicted value) It is a figure for demonstrating. It is a figure for demonstrating the copula (joint function) which concerns on the modification of this invention.

  Hereinafter, a natural energy amount prediction apparatus according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 shows a schematic configuration of a natural energy amount prediction system to which a natural energy amount prediction apparatus can be applied. Here, the amount of natural energy is the amount of energy per unit time obtained from natural energy. In this embodiment, the amount of solar radiation is predicted and provided as the amount of energy per unit time obtained from solar energy, which is natural energy. For this reason, the natural energy prediction system in this embodiment includes, for example, a photovoltaic power generation system 10 installed in a building such as a house, a management center 20 including a natural energy amount prediction device, and natural energy (more specifically, Includes a meteorological data providing center 30 that provides various meteorological data related to meteorological phenomena that affect the amount of solar radiation that is the amount of natural energy. In this natural energy amount prediction system, each photovoltaic power generation system 10, the management center 20, and the weather data providing center 30 installed in a plurality of houses or the like are, for example, a network 40 of an Internet line network or a mobile phone line network. So that they can communicate with each other.

  As shown in FIG. 2, a solar power generation system 10 installed in a house or the like is generally installed on a roof or the like on the south side of a house or the like, and is a solar cell panel 11 that converts solar energy into electrical energy (or solar A solar cell array 12 formed by connecting a plurality of battery modules 11) in series and in parallel is provided. The solar cell array 12 is connected to a power conditioner 14 through a connection box 13 having a known backflow prevention element and a DC circuit breaker. The power conditioner 14 is a so-called inverter having a conversion unit that converts DC generated power output from the solar cell array 12 (each solar cell panel 11) into AC. The power conditioner 14 is connected to a distribution board 15 connected to external system power, and the AC power converted by the power conditioner 14 is supplied to various electric devices as power consuming devices used in the home. .

  Further, the solar power generation system 10 includes a power storage device 16 as an energy storage device connected to the power conditioner 14. As shown in FIG. 2, the power storage device 16 charges, for example, a storage battery 16a such as a lithium ion battery and the electrical energy supplied via the power conditioner 14 to the storage battery 16a, and the electrical energy charged in the storage battery 16a. And a charge / discharge unit 16b for supplying electric power to various electric devices via the power conditioner 14 and the distribution board 15.

  Furthermore, the solar power generation system 10 includes a power generation monitoring device 17 connected to the power conditioner 14. The power generation monitoring device 17 acquires and monitors the amount of power generated by the solar cell array 12 (each solar cell panel 11) and cooperates with the charge / discharge unit 16b of the power storage device 16 to surplus power to the storage battery 16a. It controls the charging of the battery and the discharge when the power used is tight.

  Therefore, as shown in FIG. 3, the power generation monitoring device 17 includes an electronic control unit 17a, a communication unit 17b, a storage unit 17c, and a notification unit 17d. The electronic control unit 17a is a microcomputer whose main components are a CPU, a ROM, a RAM, and the like, and comprehensively controls the operation of the power generation monitoring device 17 by executing various programs. The communication unit 17b is connected to the network 40 to realize communication with the management center 20.

  The storage unit 17c includes a storage medium such as a hard disk and a semiconductor memory, and a drive device for the storage medium. The storage unit 17c stores in advance programs and data necessary for the electronic control unit 17a to comprehensively control the operation of the power generation monitoring device 17. Furthermore, the storage unit 17c represents the specifications of the photovoltaic power generation system 10 (specifically, the azimuth angle, inclination angle, temperature characteristics, conversion characteristics of the power conditioner 14, etc. of the solar cell array 12 (solar cell panel 11)). Specification data, regional data indicating the installation location (specifically, longitude and latitude, etc.) of the solar power generation system 10 and identification data assigned in advance for identifying the solar power generation system 10 are stored in a predetermined storage position. In addition to the storage, the power generation history data and the like in which the daily power generation amount generated by the solar power generation system 10 is stored in a time series is stored in a predetermined storage position so as to be updatable.

  The notification unit 17d includes a display display with a touch input function, a speaker, and the like. Then, the notification unit 17d displays characters, figures, etc. on the screen of the display display according to the control by the electronic control unit 17a, or outputs sound from the speaker, so that the current power generation amount and the past power generation amount will be described later. Thus, the predicted power generation amount based on the prediction information representing the predicted solar radiation amount provided from the management center 20 is notified. Needless to say, the user can appropriately change the operating state of the photovoltaic power generation system 10 using the touch input function of the display.

  The management center 20 accurately predicts and provides the amount of solar radiation that is the amount of natural energy obtained from solar energy. Therefore, the management center 20 includes a server 21 and a communication device 22 as shown in FIG.

  The server 21 includes a control device 21a, a storage device 21b, and a communication interface 21c. The control device 21a has a microcomputer composed of a CPU, ROM, RAM, etc. as a main component, and the operation of the management center 20 (more specifically, the server 21) so as to predict the amount of solar radiation that is a natural energy amount. Overall control. The storage device 21b includes a storage medium such as a hard disk and a semiconductor memory, and a drive device for the storage medium, and stores various programs and various data. The communication interface 21c is an interface for connecting to a communication line (for example, a LAN line) constructed in the management center 20. The storage device 21b includes a database 21d that stores the specification data, the regional data, and the identification data transmitted from the power generation monitoring device 17 of the photovoltaic power generation system 10 in association with each other so as to be searchable. The specification data, the regional data, and the identification data transmitted from each power generation monitoring device 17 are transmitted and stored (registered) so as to be searchable at the first communication with the management center 20, for example. It can be omitted in subsequent communications. The communication device 22 is connected to the network 40 and communicates with the power generation monitoring device 17 and the weather data providing center 30 of the solar power generation system 10.

  The meteorological data provision center 30 provides various meteorological data observed by a meteorological observation apparatus installed in the whole country. Here, in the present embodiment, as solar data that is natural energy, that is, meteorological data at each observation point related to a change (variation) in the amount of solar radiation, at least the cloud amount, wind direction, wind speed, Ascending current, temperature, atmospheric pressure and relative humidity, weather, rainfall, snowfall, snow cover, clearness, etc. are observed and provided by the weather observation device. The weather data providing center 30 provides the observed various weather data to the management center 20 via the network 40.

  Next, the function shown in FIG. 5 representing the function realized by the computer program processing in the control device 21a of the server 21 for the management center 20 configured as described above to predict the solar radiation amount that is the amount of natural energy. This will be described in detail with reference to a block diagram. The control device 21a includes an input unit 51, a first parameter calculation unit 52, a second parameter calculation unit 53, a probability density function calculation unit 54, a mode value calculation unit 55, a confidence interval calculation unit 56, an error A prediction processing unit 50 including a prediction probability calculation unit 57 and an output unit 58 is provided.

The input unit 51 as an acquisition unit inputs various information and various weather data from the power generation monitoring device 17 of the solar power generation system 10 and the weather data providing center 30. The first parameter calculation unit 52 obtains a parameter corresponding to an expected value in a probability density distribution (hereinafter also simply referred to as a probability distribution) that represents a predicted distribution of solar radiation and allows an asymmetric distribution. The second parameter calculation unit 53 as parameter calculation means obtains a parameter related to the spread in the probability density distribution that allows an asymmetric distribution. Probability density function calculation section 54 as a probability distribution calculation means calculates a probability density function that allows the asymmetric distribution to calculate the probability density distribution of the solar radiation. The mode value calculation unit 55 as the mode value calculation means calculates the mode value of the amount of solar radiation predicted based on the probability density distribution (hereinafter referred to as the predicted amount of solar radiation). A confidence interval calculation unit 56 as a confidence interval calculation means calculates a confidence interval including a mode value in the probability density distribution. The misprediction probability calculation unit 57 serving as the misprediction probability calculation means has an actual solar radiation amount that is less than the predetermined solar radiation amount outside the confidence interval (or larger than the predetermined solar radiation amount) against the predicted solar radiation amount. The probability that such a misprediction will occur is calculated. The output unit 58 outputs prediction information related to the prediction of the amount of solar radiation to the power generation monitoring device 17 of the solar power generation system 10.

Here, in the present embodiment, in order to obtain a predicted distribution of the amount of solar radiation that is the amount of energy per unit time obtained from solar energy that is natural energy, a beta distribution is adopted as a probability distribution that allows an asymmetric distribution. To do. Hereinafter, the beta distribution employed in the present embodiment will be described.

In general, the beta distribution for obtaining the distribution of the predicted value y using two positive parameters (p, q) is expressed by the following equation 1, and the probability density distribution f (y) (or probability) of the predicted value y in this case The density function f (y)) is expressed by the following equation 2.
However, Β (p, q) in Equation 2 represents a beta function using two positive parameters (p, q).

Further, the expected value E [y] and the variance Var [y] of the probability density distribution f (y) that allows the asymmetric distribution represented by the above equation 2 are generally represented by the following equations 3 and 4.

In the present embodiment, μ = p / (p + q), φ as new parameters (μ, φ) using two positive parameters (p, q) in the beta distribution generally defined as described above. = P + q is set, and a beta distribution using these parameters (μ, φ) is adopted. However, μ satisfies 0 <μ <1 and φ satisfies φ> 0. Thereby, the beta distribution in this embodiment is represented by the following formula 5, and the probability density distribution f (y) of the predicted value y is represented by the following formula 6 according to the formula 2.

Then, the expected value E [y] and the variance Var [y] of the probability density distribution f (y) represented by the equation 6 are represented by the following equations 7 and 8 according to the equations 3 and 4.
Here, as is clear from the equation 7, μ represents the expected value E [y], and as shown in the equation 8 and FIGS. 6A and 6B, φ increases as the value increases. Since y] becomes smaller, it represents the spread of the distribution.

Further, by expressing the parameters (μ, φ) as μ (x) and φ (x) as a function of the explanatory variable x related to the change (variation) of the predicted value y, the following equation 9 is obtained. The probability density distribution f (y) of the predicted value y can be obtained from the predetermined explanatory variable x.

For μ (x) and φ (x), as a function of the explanatory variable x, a linear function, a power function, an exponential function, a logit function, a probit function, a double exponential function, or the like, or Any of non-parametric values such as a numerical table (so-called map) for the explanatory variable x may be used. However, in this case, a function that takes a positive value within the range of possible values of the explanatory variable x is preferable. Specifically, parametric formulas are illustrated as shown in the following formulas 10-12. For φ (x), it is also possible to store a past calculation result history or the like and use a constant based on this calculation result. In this case, the calculation amount can be reduced, and the calculation load can be reduced.

Such explanatory variables parameters expressed as a function of x mu (x) and parameter phi (x) Beta distribution is calculated have use a (probability density distribution f (y)) adopted, the management center 20 servers 21 predicts the predicted solar radiation amount y that is the predicted value y. This will be specifically described below.

  The input unit 51 acquires, from the weather data providing center 30, various types of weather data that are targets of the explanatory variable x related to the change (variation) in the amount of solar radiation. Moreover, the input part 51 acquires area data and identification data from the electric power generation monitor apparatus 17 of each solar power generation system 10 which provides prediction information. Thereby, the input unit 51 selects various weather data in the region where the photovoltaic power generation system 10 providing the prediction information is installed based on the region data, and sets the selected various weather data as candidates for the explanatory variable x. To do. Then, the input unit 51 outputs the explanatory variable x to the first parameter calculation unit 52 and the second parameter calculation unit 53. Specifically, the input unit 51 is a cloud amount (low, middle, high) that is observed (actually measured) or predicted by a weather observation device provided in the area where the photovoltaic power generation system 10 is installed. Predictions based on wind direction, wind speed, updraft, temperature, atmospheric pressure and relative humidity, weather, precipitation, rainfall, snowfall, snow cover, clearness, etc., air mass and solar radiation, etc. The lead time up to the time is set as a candidate for the explanatory variable x, and is output to the first parameter calculation unit 52 and the second parameter calculation unit 53. In this case, in addition to the weather data in the area where the photovoltaic power generation system 10 is installed, other related time and weather data in the area can be set as candidates for the explanatory variable x.

The first parameter calculation unit 52 calculates a parameter μ (x) corresponding to an expected value in the probability density distribution f (y) represented by the equation 9. Specifically, the first parameter calculation unit 52 appropriately selects from the candidates for the explanatory variable x output from the input unit 51 to determine the explanatory variable x, and uses the explanatory variable x to set the parameter μ (x) Calculate To illustrate this, the first parameter calculation unit 52 selects and sets the middle cloud cover, the high cloud cover, the relative humidity, the rainfall, and the air mass as the explanatory variable x. Then, the parameter μ (x) which is the first parameter corresponding to the expected value is calculated. Note that the coefficients α _i , β ₀ ,..., Β _i , β _j in Equation 10 can be calculated using, for example, well-known extended beta regression.
However, x _mc in the equation 13 represents the middle cloud cover, x _hc represents the high cloud cover, x _rh represents the relative humidity, x _ra represents the rainfall, and x _airmass represents the _air mass. When the parameter μ (x) is calculated in this way, the first parameter calculation unit 52 outputs the calculated parameter μ (x) to the probability density function calculation unit 54.

The second parameter calculation unit 53 calculates a parameter φ (x) related to the distribution spread in the probability density distribution f (y) represented by the equation 9. Specifically, the second parameter calculation unit 53 determines the explanatory variable x by appropriately selecting from the candidates for the explanatory variable x output from the input unit 51, and uses the explanatory variable x to set the parameter φ (x). Calculate To illustrate this, the second parameter calculation unit 53 selects and sets a low cloud cover, a middle cloud cover, a precipitation and an air mass as the explanatory variable x. The parameter φ (x), which is the second parameter related to the spread, is calculated. The second parameter calculation unit 53 can also calculate the coefficients α _i , β ₀ ,..., Β _i , β _j in the equation 10 using, for example, a well-known extended beta regression. .
However, x _{lc in} Equation 14 represents a low cloud cover, x _mc represents a medium cloud cover, x _ra represents a rainfall amount, and x _airmass represents an _air mass.

  Here, in particular, for the explanatory variable x set in calculating the parameter φ (x) related to the spread of the distribution, for example, the weather is reflected by the specific wind direction, reflecting the characteristics of the region providing the prediction information. In an area where the phenomenon of change is likely to occur, the phenomenon that occurs can be appropriately reflected by setting the explanatory variable x with priority on the wind direction and the wind speed. Moreover, the phenomenon in which the prediction accuracy decreases with time, can be appropriately reflected in the prediction of the amount of solar radiation by including the lead time in the explanatory variable x. Further, for example, even when the predicted solar radiation amount y is separately provided, as described later, obtaining the parameter φ (x) corresponding to the spread of the distribution accurately results in misprediction (outlier). This is also useful for predicting well, and in this case, the predicted solar radiation amount y provided for the explanatory variable x can also be set. When the parameter φ (x) is calculated in this way, the second parameter calculation unit 53 outputs the calculated parameter φ (x) to the probability density function calculation unit 54.

The probability density function calculation unit 54 relates to the parameter μ (x) corresponding to the expected value calculated by the first parameter calculation unit 52 and the spread of the distribution calculated by the second parameter calculation unit 53 according to Equation 9. The probability density distribution f (y) is calculated using the parameter φ (x). Here, the amount of solar radiation to be predicted is a value in a finite section with “0” as the lower limit and the amount of solar radiation outside the atmosphere as the upper limit. Therefore, the prediction error of the predicted solar radiation amount y (predicted value) is It follows an asymmetric probability distribution. In this regard, in the probability density distribution f (y) calculated using the parameter μ (x) and the parameter φ (x) in accordance with the equation 9, the error distribution of the predicted solar radiation amount y is as shown in FIG. An asymmetric distribution. Further, in the probability density distribution f (y) that allows an asymmetric distribution, as is apparent from FIG. 7, the mode value y _mode appears separately for the parameter μ (x), that is, the expected value μ (x). To do. That is, as indicated by a solid line in FIG. 8, in the probability density distribution f (y) (beta distribution) that allows an asymmetric distribution, the expected value μ (x) and the mode y _mode exist individually. On the other hand, in the method of obtaining the expected value μ (x) based on the normal distribution, such as the least square method, the expected value μ (x), the mode y _mode , as indicated by the broken line in FIG. And the median values are all the same, and the expected value μ (x) and the mode value y _mode cannot be distinguished.

  Furthermore, by expressing the predicted solar radiation amount y with a probability density distribution f (y) (beta distribution) that allows an asymmetric distribution, the shape of the tail of the distribution can be accurately expressed. Specifically, in the probability density distribution f (y) calculated according to Equation 9, the parameter φ (x) related to the spread of the distribution is calculated as a function of the explanatory variable x. Can be accurately matched to reality. Then, the probability density function calculation unit 54 outputs the calculated probability density distribution f (y) to the mode value calculation unit 55, the confidence interval calculation unit 56, and the misprediction probability calculation unit 57.

Mode value computing unit 55, modeled using the probability allowing the calculated asymmetric distribution by a probability density function calculation unit 54 density distribution f (y) (Beta distribution), the probability density distribution f (y) Thus, the mode value y _mode that appears as shown in FIG.
However, μ (x) in the equation 15 is a parameter μ (x) calculated as a function of the explanatory variable x by the first parameter calculator 52, and φ (x) in the equation 15 is a second parameter calculation. This is a parameter φ (x) calculated by the unit 53 as a function of the explanatory variable x.

Here, in the natural energy amount prediction apparatus, at least the mode y _mode is used as the predicted solar radiation amount y. As a result, it is possible to accurately represent the change in the amount of solar radiation (the likelihood of occurrence of an event) and provide a highly accurate predicted amount of solar radiation y. This will be described with reference to FIGS.

  FIG. 9 shows an example of prediction when the distribution of the predicted solar radiation amount y has a prediction error according to a normal distribution that is a symmetrical distribution of infinite intervals. In this case, since the probability density distribution indicated by the dotted line in FIG. 9 has a prediction error according to the normal distribution in the infinite interval, a situation different from the reality that the lower limit of the prediction error becomes negative occurs. That is, in this case, since the shape of the tail of the probability density distribution is not determined so as to accurately reflect the actual situation, it is extremely difficult to accurately calculate the probability of misprediction (outlier) described later. . In this case, since the distribution is symmetric, the mode value and the expected value cannot be distinguished. Therefore, in this case, as shown by the long broken line in FIG. 9, the expected value of the probability density distribution is the predicted solar radiation amount y (predicted value), and the actual measured value of the solar radiation amount shown by the solid line in FIG. There is a portion (for example, a time zone) that is significantly different from that (hereinafter also referred to as the actually measured solar radiation amount).

Figure 10 shows predicted example in which the distribution of the predicted insolation y has to have a prediction error according Beta distribution is a distribution to allow and asymmetric finite interval. In this case, since the probability density distribution f (y) indicated by the dotted line in FIG. 10 has a prediction error according to the beta distribution, it accurately reflects the actual situation where the prediction error is always “0” or more and below the atmospheric solar radiation. can do. In this way, since the shape of the tail of the probability density distribution f (y) is accurately expressed and is determined to reflect the actual situation, the probability of misprediction (outlier) can be accurately calculated. . In this case, since the beta distribution allows asymmetry, the mode value y _mode and the expected value μ (x) can be distinguished. Therefore, as shown by a chain line in FIG. 10, you are possible to predict solar radiation the mode y _mode of the probability density distribution f (y) y (predicted value). Then, the predicted solar radiation amount y (predicted value), which is the mode y _mode of the probability density distribution f (y), is the predicted solar radiation amount y (predicted value) of the expected value μ (x) of the probability density distribution f (y). Compared with the case (indicated by the long broken line in FIG. 10), the measured solar radiation amount indicated by the solid line in FIG. 10 is closer.

  In this regard, FIG. 11 shows the relationship between the predicted solar radiation amount y (predicted value) and the actually measured solar radiation amount (actually measured value). In FIG. 11, the predicted amount of solar radiation y is normalized by the amount of solar radiation outside the atmosphere, and the amount of actually measured solar radiation is normalized by the amount of solar radiation outside the atmosphere.

FIG. 11A shows a case where the expected value μ (x) is used as the predicted solar radiation amount y (predicted value). In this case, the expected value μ (x) is calculated and used in the probability density distribution f (y) (beta distribution), which is an asymmetric distribution. The quantity y is underestimated. As a result, as shown in FIG. 11A, the relationship between the predicted solar radiation amount y and the actually measured solar radiation amount is clearly non-linear (S-shaped). On the other hand, FIG. 11B shows a case where the mode y _mode is used as the predicted solar radiation amount y (predicted value). In this case, as shown in FIG. 11B, the relationship between the predicted solar radiation amount y and the actually measured solar radiation amount becomes more linear. Therefore, it can be understood that the accuracy of prediction is improved by using the mode y _mode as the predicted solar radiation amount y. As described above, when the mode value y _mode is calculated, the mode value calculation unit 55 outputs the _mode value y _mode to the output unit 58.

Confidence interval calculation unit 56, the probability allowing the calculated asymmetric distribution by a probability density function calculation unit 54 density distribution f (y) (beta distribution), as shown in FIG. 12, includes a mode value y _mode the confidence interval Y _a, is determined by calculating a set of prediction insolation y satisfying the following equation 16.
However, a in the equation 16 is a probability density distribution f (y), that is, a probability density distribution f (y _made ) in accordance with the probability density distribution f (y _made ) in the maximum mode y _mode among the frequencies f (y). This is a variable for determining the size of y).

Will be specifically described below confidence interval Y _a is determined according to the equation 16. As described above, the mode value calculation unit 55 calculates the _mode value y _mode by modeling using the probability density distribution f (y) (beta distribution) calculated by the probability density function calculation unit 54. This mode value y _mode can be used as the predicted solar radiation amount y. As a result, it is possible to provide a prediction closer to the actual situation when the _mode value y _mode is the predicted solar radiation amount y than when the expected value μ (x) is the predicted solar radiation amount y.

By the way, when the expected value μ (x) or the median value is used for the predicted solar radiation y (predicted value) as in the case of a normal distribution having a symmetric distribution, usually, for example, the standard deviation or the percentile value is It is given as a confidence interval for the predicted solar radiation y (predicted value) (equal posterior distribution interval). However, in the probability density distribution f (y) (beta distribution) that allows an asymmetric distribution, as described above, when the mode y _mode is used as the predicted solar radiation amount y (predicted value), it is shown in FIG. As described above, a contradiction may occur such that the mode y _mode (predicted solar radiation amount y) does not enter (is not included in) the confidence interval given as the above-mentioned equi-tailed posterior distribution interval.

Therefore, confidence interval calculation unit 56, as determined according to the equation 16, the probability density distribution f (y) i.e. the frequency f (y) is at its maximum value f (y _mode) a large value in accordance with af ( the highest posterior density interval is a set of predicted insolation y (predicted value) that satisfies y _mode) is determined as a confidence interval Y _a. In other words, the confidence interval Y _a is determined in this way, as is apparent from FIG. 12, it corresponds to considering the contours of the frequency f (y) (the probability density distribution f (y)). Thus, the mode y _mode is always (included) enters the confidence interval Y _a, it is possible to avoid the inconsistency described above occurs. Even if the predicted solar radiation amount y (predicted value) that takes a sufficiently large value according to the maximum value f (y _mode ) is adopted, the same effect as the mode y _mode described above is used. Can be obtained.

Moreover, by determining the thus confidence interval Y _a, it can be determined properly reflect the confidence interval real situation. That is, for example, when a confidence interval based on the expected value μ (x) or the median value is determined assuming that the error distribution of the predicted solar radiation amount y (predicted value) follows a normal distribution of an infinite interval, the upper diagram of FIG. As shown in FIG. 6, a situation different from the reality that the confidence interval becomes negative may occur. In contrast, when determining the confidence interval Y _a error distribution is based on the most frequent value y _mode as follow beta distribution to permit an asymmetric distribution of the predicted insolation y (predicted value) of FIG. 14 as shown below, becomes confidence interval Y _a is always "0" or more, the reality of the situation can be properly reflected. In FIG. 14, the dark gray area indicates the 50% confidence interval of the frequency f (y) (probability density distribution f (y)), and the light gray area indicates the frequency f (y) (probability density distribution f (y). )) 90% confidence interval. In FIG. 14, the solid line indicates the actual amount of solar radiation, the long broken line in the upper diagram of FIG. 14 indicates the expected value μ (x) as the predicted value y, and the long broken line in the lower diagram of FIG. 14 indicates the mode value. The y _mode is the predicted value y.

Here, as shown in FIG. 12, the area obtained by integrating the frequency f (y) a (probability density distribution f (y)) with a confidence interval Y _a is reliability Pr (Y in this confidence interval Y _a It corresponds to _a ). That is, the reliability Pr (Y _a ) can be calculated according to the following equation 17.
It is also possible to calculate the desired confidence variable a from P reversed and confidence interval Y _a. In this case, for example, when the probability density function of a unimodal probability distribution is set to f (y) and the cumulative distribution function is set to F (y), by solving the simultaneous equations expressed by the following Equation 18, A confidence interval Y = [y _n , y _m ] with confidence P around the value can be obtained.
Thus, when calculating the confidence interval Y _a, confidence interval calculation unit 56 outputs the confidence interval Y _a to the output unit 58.

In the probability density distribution f (y) (beta distribution) that allows the asymmetric distribution calculated by the probability density function calculator 54, the misprediction probability calculator 57 appears near the tail of the distribution as shown in FIG. The lower limit side probability Pr_e (y <y _L ) and the upper limit side probability Pr_e (y> y _H ) of the misprediction (large deviation) are calculated according to the following equation 19. In the following description, the lower limit probability Pr_e (y <y _L ) and the upper limit probability Pr_e (y> y _H ) are collectively referred to as an erroneous prediction probability Pr_e.
However, as shown in FIG. 12, y _L in the equation 19 represents a lower limit predicted value preset in the probability density distribution f (y) determined according to the equation 9, and y _H represents the equation 9 Represents an upper limit predicted value set in advance in the probability density distribution f (y) determined according to.

Here, the misprediction probability calculation unit 57 uses the probability density distribution f (y) calculated by the probability density function calculation unit 54 in accordance with the equation 19, and uses the lower limit probability Pr_e (y <y _L ) and the upper limit probability. Calculate Pr_e (y> y _H ). As a result, as shown in FIG. 10, the lower limit side of the prediction error is always “0” or more, and the upper limit side of the prediction error can be less than or equal to the amount of solar radiation outside the atmosphere. ) can be very compute accurately the lower side probability Pr_e generated (y <y _L) and the upper side probability Pr_e (y> y _H). That is, in order to determine the probability density distribution f (y), the second parameter calculation unit 53 calculates the parameter φ (x) related to the distribution expansion as a function of the explanatory variable x. Thus, the shape of the tail of the probability density distribution f (y) that allows an asymmetric distribution can be accurately expressed. Therefore, the misprediction probability calculation unit 57 calculates the lower limit probability Pr_e (y <y _L ) and the upper limit probability Pr_e (y> y _H ) in the vicinity of the tail of the probability density distribution f (y) according to the above equation 19. It can be calculated and determined with high accuracy. Then, the misprediction probability calculation unit 57 calculates the misprediction probability Pr_e of the misprediction (outlier) and outputs it to the output unit 58.

The output unit 58 includes the mode y _mode supplied from the mode calculation unit 55, the confidence interval Y _a supplied from the confidence interval calculation unit 56, and the misprediction probability supplied from the misprediction probability calculation unit 57. Pr_e (specifically, lower limit probability Pr_e (y <y _L ) and upper limit probability Pr_e (y> y _H )) is acquired. Then, the output unit 58 outputs, as prediction information, to the power generation monitoring device 17 of the solar power generation system 10 installed in the area where the amount of solar radiation is predicted via the communication device 22 and the network 40. prediction insolation information representing the prediction insolation y is Shikichi y _mode, and the confidence interval information indicating the confidence interval Y _a, the lower limit probability Pr_e (y <y _L) and the upper side probability Pr_e (y> y _H) And misprediction probability information representing.

Here, in the power generation monitoring device 17 that has received the predicted solar radiation amount information, the confidence interval information, and the erroneous prediction probability information, the power generation amount by the solar power generation system 10 is calculated and predicted. That is, the electronic control unit 17a of the power generation monitoring device 17 acquires the predicted solar radiation amount information, the confidence interval information, and the misprediction probability information via the communication unit 17b, and the acquired information is stored in a predetermined storage position of the storage unit 17c. To remember. Then, the electronic control unit 17a determines the installation state (azimuth angle, inclination angle, etc.) of the solar cell array 12 (solar cell panel 11) of the photovoltaic power generation system 10 represented by the specification data, the conversion characteristics of the power conditioner 14, and the like. Using the predicted solar radiation amount y (more specifically, the mode y _mode ) represented by the predicted solar radiation information and the confidence interval Y _a represented by the confidence interval information, that is, the horizontal solar radiation amount. Calculate the amount of power generation that is sometimes predicted. The electronic control unit 17a notifies the user of the predicted power generation amount, for example, via the notification unit 17d.

  On the other hand, the electronic control unit 17a of the power generation monitoring device 17 determines the power generation amount of the power storage device 16 when the power generation amount is predicted to increase, for example, according to the power generation amount by the solar power generation system 10 calculated and predicted. In cooperation with the charging / discharging unit 16b, the surplus power is charged in advance of charging using the night power for the storage battery 16a, or the power of the storage battery 16a is discharged to sell the surplus power. Conversely, when it is predicted that the amount of power generated by the solar power generation system 10 will be small, the electronic control unit 17a cooperates with the charge / discharge unit 16b to, for example, actively use nighttime power to store the storage battery. Inexpensive power is charged in advance in 16a, or power consumption of various electric devices connected to the distribution board 15 is suppressed.

Further, the electronic control unit 17a of the power generation monitoring device 17 makes the provided predicted solar radiation amount y (specifically, the mode value y _mode ) an erroneous prediction based on the erroneous prediction probability information stored in the predetermined storage position. When the probability (so-called great deviation) is relatively high, the charge amount of the storage battery 16a can be adjusted in cooperation with the charge / discharge unit 16b of the power storage device 16. Specifically, for example, although the predicted solar radiation amount y (mode value y _mode ) is predicted to increase, the actually measured solar radiation amount is “0” for several times in a year (several days). ”May occur. On the other hand, for example, although the predicted solar radiation amount y (mode y _mode ) is predicted to be “0”, the actual solar radiation amount increases several times (several days) in one year. A misprediction (outlier) can occur.

When such a misprediction (outlier) occurs, the user of the photovoltaic power generation system 10 must not use a commercial power supply because the storage battery 16a of the power storage device 16 is not sufficiently charged. It becomes a situation that does not become a situation, or a situation occurs where the surplus power cannot be sold efficiently. For this reason, when the electronic control unit 17a has the lower limit probability Pr_e (y <y _L ) and the upper limit probability Pr_e (y> y _H ) represented by the misprediction probability information larger than the preset probability Pr_err. Specifically, when the lower limit probability Pr_e (y <y _L ) is greater than the probability Pr_err, the predicted solar radiation amount y (mode value y _mode ) may become “0”. 16a is charged with nighttime power in advance. On the contrary, when the upper limit probability Pr_e (y> y _H ) is larger than the probability Pr_err, there is a possibility that the predicted solar radiation amount y (mode y _mode ) of “0” may increase. A capacity that can be charged is provided in the storage battery 16a.

  As can be understood from the above description, according to the above embodiment, the parameter φ (x) related to the spread of the distribution is calculated using the explanatory variable x, and the probability density distribution using the parameter φ (x) is calculated. Using f (y), it is possible to calculate the predicted solar radiation amount y of the solar radiation amount, which is the natural energy amount per unit time obtained from natural energy, including the prediction error distribution with extremely high accuracy. Further, the misprediction probability Pr_e in which a rare phenomenon such that the predicted solar radiation amount y is erroneously predicted (the predicted solar radiation amount y greatly deviates) can be predicted quantitatively and accurately.

<First Modification>
In the above embodiment, the prediction processing unit 50 is performed to adopt the beta distribution as the probability density distribution to allow asymmetrical. In this case, as an asymmetric probability distribution that handles positive continuous values such as solar radiation and power generation related to this solar radiation, in addition to the beta distribution, exponential distribution family, lognormal distribution, gamma distribution, inverse gamma distribution, Inverse Gaussian distribution, Tweedie distribution, Weibull distribution, Gumbel distribution, and the like can be given. For this reason, it is also possible to carry out the modification by using these probability distributions. That is, in each of these probability distributions, there is a parameter closely related to the spread of the distribution as in the case of the beta distribution. Therefore, in each of the probability distributions, the parameters related to the distribution spread are expressed as a function of explanatory variables, so that the distribution of the predicted value distribution, the mode value based on the frequency, and the mode based on the frequency can be obtained. It is possible to calculate a confidence interval of a value, a probability that an erroneous prediction (outlier) occurs, and the like.

Specifically, when an exponential distribution family is used as the asymmetric probability distribution for positive continuous values, the probability density function can be expressed by the following Equation 20.
However, φ in the equation 20 is a parameter related to the spread of the distribution, and the parameter φ representing the spread of the distribution by expressing the parameter φ as a function of the explanatory variable x as in the above-described embodiment. x) can be calculated. Therefore, also in this first modified example, the same effect as in the above embodiment can be obtained.

<Second Modification>
In the embodiment and the first modification, at least, the predicted insolation y at a certain point in time (a time zone) (most frequent value y _mode), the lower limit confidence interval Y _a and erroneous predictions (utter failure) occurs Probability Pr_e (y <y _L ) and upper limit probability Pr_e (y> y _H ) were calculated. In this case, it is also possible to update (correct) the predicted value (predicted solar radiation amount y) and the predicted distribution in real time by using the actual value of the weather data or the latest observed weather data for the explanatory variable x. It is. In this case, for example, a correlation between prediction distributions of prediction values (predicted solar radiation amount y) between different times can be expressed using a joint function (copula), and a conditional probability can be calculated. Hereinafter, this second modification will be specifically described.

A copula (joint function) is a function that gives a relationship between two or more arbitrary random variables. Thus, a copula can be applied between variables that follow, for example, two beta distributions. Specifically, the copula of the two-dimensional data (y ₁ , y ₂ ) is generally a function called C (U ₁ , U ₂ ), and the peripheral distribution of each dimension is F ₁ (y ₁ ). , F ₂ (y ₂ ), the two-dimensional distribution is expressed by the following equation (21).

Here, exemplarily, for example, in the two-dimensional data (y ₁ , y ₂ ), y ₁ is a variable representing the amount of solar radiation at 9 o'clock, and y ₂ is 10 o'clock in the range from 9 o'clock. Assuming that it is a variable representing the amount of solar radiation, F (y ₁ , y ₂ ) represents the simultaneous distribution. Therefore, when the measured value of the solar radiation amount y ₁ at 9 o'clock is given, the solar radiation amount at 10 o'clock ₂ conditional density function f (y ₂ | y ₁₎ can be calculated according to the following equation 22.
Therefore, according to the above equation 22, it is possible to calculate the predicted value (estimated solar radiation y in the 10 o'clock range) with accuracy guarantee that reflects the measured value (actual solar radiation amount in the 9 o'clock range) in real time, that is, corrected in real time. It becomes.

Here, in general, there are many types of copulas such as a Kraton type, a Frank type, a Gumbel type, a Gauss type, and an Archimedes type. Among these, there are many copulas that can be expressed by one parameter θ, and in the case of two dimensions, the Kraton type, the Gauss type, the copulas of the above formulas 21 and 22, and the like correspond to them. And, for example, such a copula, for example, a Clayton copula can be expressed as the following Expression 23.
Thus, in the copula that can be represented by one parameter θ, for example, the correlation becomes stronger as the parameter θ increases as a positive value, and the correlation becomes uncorrelated when the parameter θ is approximately “0”. If the parameter θ is a negative value, a negative correlation is obtained.

  Specifically, the correlation between the above 9 o'clock solar radiation amount and the 10 o'clock solar radiation amount will be described as an example. For example, in the case of sunny weather, as shown in FIG. It is assumed that the correlation between 9 o'clock and 10 o'clock (consecutive) becomes strong, in other words, the parameter θ increases as a positive value. On the other hand, in the weather in which an increase in clouds or rainfall is expected, as shown in FIG. 15B, the correlation between the adjacent (continuous) 9 o'clock and 10 o'clock is weak, in other words, the parameter θ is “ It is assumed that it approaches “0”. In other words, when predicting the amount of solar radiation, it can be expected that the correlation between adjacent (continuous) time zones is strong in well-clear weather. Then, it is only necessary to assume that the parameter θ approaches “0” because the correlation between adjacent (continuous) time zones becomes weak.

Therefore, the second parameter calculation unit 53 similarly indicates that the parameter φ related to the spread of the distribution in the above-described embodiment is expressed as the function φ (x) of the explanatory variable x between adjacent (continuous) time zones. By expressing the correlation as a function of the explanatory variable x called the parameter θ (x), the correlation between the time zones can be appropriately expressed. In this case, as the explanatory variable x, for example, the actual value of each meteorological data provided from the meteorological data providing center 30, each recently observed meteorological data, or the like can be used. And by using the parameter θ (x) in this way, the predicted value (predicted solar radiation amount y) at each time point (each time zone) can follow the beta distribution that allows asymmetry as described above, And, using the measured value (actual solar radiation amount) in the time zone (previous time zone) that includes the current time in real time to correct the predicted value (predicted solar radiation amount y) in the next time zone (later time zone) It can be reflected.

For example, by using the parameter θ (x) calculated according to the following equation 24, when the cloud amount x _mc is “0”, the parameter θ (x) becomes a positive value and the correlation becomes strong, and the cloud amount x _mc becomes “ When “1”, the parameter θ (x) becomes “0”, and it can be expressed that the correlation is weakened.
The coefficient e in the equation 24 can be determined based on various data.

  Therefore, by using the copula (joint function), the intensity of correlation between consecutive time zones (time correlation) is taken into account, and the measured solar radiation amount (actual measurement value) in the previous time zone is estimated as the predicted solar radiation in the later time zone. It can be reflected in the quantity y (predicted value). That is, when the parameter θ (x) is a positive value and the correlation is strong, the predicted solar radiation amount y (predicted value) in the later time zone is used as the measured solar radiation amount (actually measured value) in the previous time zone. It can correct | amend suitably and the prediction precision of the predicted solar radiation amount y (predicted value) can be improved significantly. Therefore, in the second modified example, since the predicted solar radiation amount y (predicted value) with improved prediction accuracy can be used, the same effect as in the above embodiment can be expected.

  In this case, the parameter θ may be set as a fixed value without being set as a function of the explanatory variable x. In this case, not only the time correlation but also the correlation between adjacent points can be expressed by a copula and the parameter θ at this time can be set as a function of the explanatory variable x. Furthermore, in this case, not only continuous (adjacent) temporal correlations, but also the correlation between different time points is represented by a copula, and the predicted value (prediction) at a later time point is calculated using the actual value (actual solar radiation amount) at the previous time point It is also possible to implement so as to correct the amount of solar radiation y).

  In carrying out the present invention, the present invention is not limited to the embodiment, the first modified example, and the second modified example, and various modifications can be made without departing from the object of the present invention.

  For example, in the above-described embodiment and each modification, the well-known extended data regression is performed for the parameter μ (x) calculated by the first parameter calculation unit 52 and the parameter φ (x) calculated by the second parameter calculation unit 3. It was carried out to estimate and calculate at the same time using In this case, as a simpler method of calculating the parameter μ (x) and the parameter φ (x), first, assuming a normal distribution that is a symmetric distribution, the parameter μ (x) is determined by the least square method, and the parameter It is also possible to maximize the likelihood of the parameter φ (x) after determining μ (x). Also, for example, when calculating multi-point predicted solar radiation y such as 24 hours or 12 hours per day, it is assumed that they follow a multidimensional normal distribution, and statistically considering the explanatory variable x It is also possible to calculate the parameter φ (x) related to the spread of the distribution by using the parameter μ (x) as the parameter μ (x) obtained from the average value vector of the distribution.

  In addition, as a method of calculating the parameter μ (x) and the parameter φ (x), for example, the correlation between the solar radiation amounts at different times is obtained using the copula (joining function) described above, and the average is calculated using this correlation. It is also possible to calculate the parameter φ (x) related to the spread of the distribution by using the parameter μ (x) as the parameter μ (x) obtained from the value vector. Furthermore, when the predicted solar radiation amount y is provided separately, the provided predicted solar radiation amount y can be used as the parameter μ (x), and then the parameter φ (x) can be calculated.

  Moreover, in the said embodiment and each modification, it implemented so that the natural energy amount prediction apparatus might estimate the solar radiation amount as an energy amount obtained from sunlight which is natural energy. In this case, the natural energy amount prediction device predicts the amount of energy per unit time obtained from natural energy that changes (fluctuates) due to weather phenomena (meteorological conditions), such as solar thermal energy, wind energy, tidal current energy, and water current energy. It is also possible to do. The amount of energy obtained from such solar thermal energy, wind energy, tidal current energy, water current energy, and the like also changes (fluctuates) due to weather phenomena (meteorological conditions), similar to the above-described sunlight. For this reason, similarly to the above-described embodiment and each modification, it is possible to accurately predict the energy amount that is the predicted value including the distribution of the prediction error, and the predicted value is erroneously predicted (outlier). Rare phenomena can also be predicted quantitatively.

  Moreover, in the said embodiment and each modification, it implemented so that the natural energy amount prediction apparatus might predict the solar radiation amount (predicted value) which is the energy amount per unit time obtained from sunlight energy. In this case, it is also possible for the natural energy amount prediction apparatus to calculate and provide a predicted distribution of power generation amount and a predicted value of power generation amount based on the specification data acquired from each solar power generation system 10. . In this case, the amount of conversion by which the natural energy amount prediction device converts the amount of energy obtained from the natural energy (for example, solar radiation, solar heat, wind power, tidal current and water flow) into other energy (for example, electric energy). Needless to say, it is possible to predict this. Thereby, in the power generation system that converts the amount of energy obtained from natural energy into electric energy, the natural energy amount prediction device predicts the power generation amount with high accuracy, for example, so that other power generation systems, energy storage devices, The necessary electrical energy can be efficiently and reliably ensured in cooperation with the power consuming device.

In the above-described embodiment and each modification, the beta distribution in which the probability of occurrence of “0” or “1” is normally “0” is adopted. In this case, the zero-inflated bata distribution and the “1” one-inflated bata (one-inflated bata distribution) improved from the point that the probability of occurrence of “0” or “1” is “0”.
It is also possible to carry out so as to predict the amount of solar radiation using distribution). It is also possible to adopt a Dirichlet distribution that expands the beta distribution to multivariate, such as a multidimensional normal distribution that expands the normal distribution to multivariate, so that the amount of solar radiation in multiple time zones can be predicted simultaneously. It is.

  Furthermore, in the said embodiment and each modification, the management center 20 provided and implemented the natural energy amount prediction apparatus. In this case, it is needless to say that the natural energy amount prediction device can be implemented, for example, in a house where the solar power generation system 10 is installed. In this case, for example, a natural energy amount predicting device can be provided in the computer device or the power generation monitoring device 17 provided in the house.

  DESCRIPTION OF SYMBOLS 10 ... Solar power generation system, 11 ... Solar cell panel, 12 ... Solar cell array, 13 ... Connection box, 14 ... Power conditioner, 15 ... Distribution board, 16 ... Power storage device, 16a ... Storage battery, 16b ... Charge / discharge unit, DESCRIPTION OF SYMBOLS 17 ... Power generation monitoring apparatus, 20 ... Management center, 21 ... Server, 21a ... Control device, 22 ... Communication device, 30 ... Weather data provision center, 40 ... Network, 50 ... Prediction processing part, 51 ... Input part, 52 ... First 1 parameter calculator, 53 ... second parameter calculator, 54 ... probability density function calculator, 55 ... mode value calculator, 56 ... confidence interval calculator, 57 ... misprediction probability calculator, 58 ... output unit

Claims (11)

A power generation system that converts energy obtained from natural energy that changes due to meteorological phenomena into electrical energy, supplies it to a power consuming device, and charges a storage battery. The amount of natural energy per unit time obtained from the natural energy A power generation system that controls charging and discharging of the storage battery using the prediction information of
Applied to a natural energy prediction system comprising a meteorological data providing center for providing meteorological data related to meteorological phenomena affecting the natural energy,
In the natural energy amount prediction device comprising a predicting unit that is communicably connected to the power generation system and the weather data providing center and predicts the natural energy amount using the weather data from the weather data providing center ,
Wherein are those predicted distribution of natural energy is represented by the asymmetric probability distribution,
The prediction means includes
Obtaining weather data in the area where the power generation system is installed as an explanatory variable related to the change in the amount of natural energy from the weather data providing center,
Using the acquired explanatory variable, a first parameter corresponding to an expected value of the probability distribution and a second parameter related to the spread of the probability distribution are calculated,
Calculating the probability distribution representing the predicted distribution of the natural energy amount using the calculated first parameter and second parameter;
Based on the calculated probability distribution, predict the predicted value of the natural energy amount,
A natural energy amount prediction apparatus that outputs the predicted value of the predicted natural energy amount to the power generation system as prediction information of the natural energy.   The power generation system is applied to a natural energy prediction system that predicts a power generation amount based on the prediction information of the natural energy amount, and controls charging / discharging of the storage battery according to the predicted power generation amount. Natural energy amount prediction device. In the natural energy amount prediction apparatus according to claim 1 or 2,
The prediction means includes
Calculating the mode value that appears most frequently in the calculated probability distribution;
A natural value characterized in that at least the mode value of the calculated mode value and a value that frequently appears in accordance with the mode value in the probability distribution is used as a predicted value of the natural energy amount. Energy amount prediction device. In the natural energy amount prediction device according to any one of claims 1 to 3,
The prediction means includes
Determining a confidence interval of the predicted value of the natural energy amount, which is a set of values including at least the mode value that appears most frequently in the calculated probability distribution and appearing at a predetermined frequency in the probability distribution; ,
The determined confidence interval is included in the natural energy amount prediction information and is output to the power generation system. In the natural energy amount prediction device according to any one of claims 1 to 4,
The prediction means includes
Based on the calculated probability distribution, the measured actual natural energy amount is outside the confidence interval of the predicted value of the natural energy amount and contrary to the predicted value of the natural energy amount. Calculate the probability of occurrence of a misprediction that is less than the amount of natural energy, or that the actual amount of natural energy is greater than or equal to a preset upper limit amount of natural energy,
The natural energy amount prediction apparatus including the calculated probability of occurrence of misprediction in the natural energy amount prediction information and outputting the information to the power generation system. In the natural energy amount prediction apparatus according to any one of claims 1 to 5,
The prediction means includes
A third parameter representing a correlation between the predicted distributions of the natural energy amount at different time points is calculated using explanatory variables acquired corresponding to the different time points,
Using the calculated third parameter, calculate a joint function that gives a correlation between the predicted distributions of the natural energy amount at the different time points,
When a correlation is recognized between the predicted distributions of the natural energy amount at the different time points according to the calculated third parameter, the predicted distribution of the natural energy amounts at the different time points is calculated using the calculated junction function. A natural energy characterized in that an actual natural energy amount actually measured corresponding to a predicted distribution of the natural energy amount at a previous time point is reflected in a predicted value of the natural energy amount predicted at a later time point. Quantity prediction device. In the natural energy amount prediction device according to any one of claims 1 to 6,
The prediction means includes
As the explanatory variable,
Temperature, atmospheric pressure, humidity, precipitation, precipitation, snowfall, snowfall, snow cover, low cloud cover, middle cloud cover, high cloud cover, wind direction, wind speed, upflow, weather, clearness, air mass, and solar radiation A natural energy amount prediction apparatus that acquires and uses at least one of a value and a lead time from a prediction execution time to a prediction target time. In the natural energy amount prediction device according to claim 7,
The prediction means includes
As the explanatory variable,
A natural energy amount prediction apparatus that acquires and uses at least one of the weather forecast values, the lead time, and the present conditions and weather observation values observed in the past. In the natural energy amount prediction device according to any one of claims 1 to 8,
The asymmetric probability represents the predictive distribution of the natural energy distribution,
A natural energy amount prediction apparatus characterized by being a beta distribution or a zero excess beta distribution. In the natural energy amount prediction apparatus according to any one of claims 1 to 9,
The natural energy is
A natural energy amount predicting device characterized by being at least one natural energy among solar energy, solar thermal energy, wind energy, tidal current energy and water current energy. In the natural energy amount prediction apparatus according to claim 10,
When the amount of natural energy is the amount of solar radiation per unit time obtained from the solar energy,
The prediction means includes
Using the predicted distribution of the solar radiation amount represented by the probability distribution and the predicted value of the solar radiation amount predicted based on the probability distribution,
Predicting the power generation amount distribution and power generation amount of the solar power generation system that converts the solar energy into electric energy,
A natural energy amount prediction device that coordinates the photovoltaic power generation system with another power generation system, an energy storage device, and a power consuming device based on the predicted power generation amount distribution and power generation amount.