Method for predicting generated power, device for predicting generated power, and solar power generation system

Abstract

[Problem] The present invention addresses the problem of providing a method for predicting solar-generated power, a device for predicting solar-generated power, and a solar power generation system with which it is possible to predict the power generated at a desired date and time. [Solution] This method for predicting solar power generation includes executing, in order: a first step for saving past data in which weather data and solar cell power generation data outputted via a PCS at least one year before a date subject to prediction are associated with one another; a second step for calculating a clear-weather power generation curve from the past data; a third step for obtaining a yearly-transition curve indicating the yearly transition of the total amount of power generated; a fourth step for obtaining a clear-weather power generation model from the clear-weather power generation curve in conformance with the yearly-transition curve; a fifth step for obtaining a clear-weather power generation model for the entirety of one year; a sixth step for obtaining a weather coefficient; and a seventh step for acquiring a weather forecast and obtaining predicted power generation from the clear-weather power generation model and the weather coefficient.

Description

Power generation prediction method, power generation prediction device, and solar power generation system

[0001] The present invention relates to a photovoltaic power generation power generation prediction method, a photovoltaic power generation power generation prediction device, and a photovoltaic power generation system.

[0002] Photovoltaic power generation systems that generate power using sunlight as a renewable energy source have been known for some time. However, the amount of power generated by photovoltaic power generation is affected by weather conditions. Therefore, in order to efficiently use the power generated by sunlight, it is necessary to predict the amount of power generated. For example, Patent Documents 1 and 2 disclose methods for predicting photovoltaic power generation.

[0003] JP 2017-127140 A JP 2012-124188 A

[0004] Conventional methods for predicting power generation are based on solar radiation, and require a correlation between actual measured values ​​of solar radiation and actual measured values ​​of power generation. However, the actual measured values ​​of solar radiation available on the Internet are limited, and a solar radiation meter must be installed to obtain accurate measured values ​​of solar radiation at the location where the solar cells are installed.

[0005] An object of the present invention is to provide a method for predicting photovoltaic power generation power, a photovoltaic power generation prediction device, and a photovoltaic power generation system that are capable of predicting power generation at a desired date and time in a photovoltaic power generation system without requiring a solar radiation meter.

[0006] The photovoltaic power generation system refers to a power generation system equipped with a solar cell and a power conditioner.

[0007] The method for predicting power generation according to the present invention is a method for predicting power generation for a solar power generation system including a solar cell and a power conditioner, and is characterized by executing the following steps in this order: a first step of storing past data correlating power generation data of the solar cell output through the power conditioner with weather data for at least one year prior to the target day for power generation prediction; a second step of obtaining a clear-sky power generation curve from the power generation values ​​of a day selected from the past data; a third step of obtaining an annual trend curve showing the annual trend of total power generation obtained by integrating the clear-sky power generation curve; a fourth step of scaling the clear-sky power generation curve to match the annual trend curve to obtain a clear-sky power generation model; a fifth step of obtaining the clear-sky power generation model for all days in a year; a sixth step of obtaining weather coefficients according to each weather; and a seventh step of obtaining a weather forecast for the target day for prediction and predicting the power generation output from the power conditioner from the clear-sky power generation model and the weather coefficients for the target day for prediction.

[0008] By using such a method for predicting power generation, it becomes possible to predict power generation by photovoltaic power generation without the need for a solar radiation meter.

[0009] In addition, the method for predicting power generation according to the present invention is characterized in that the clear weather power generation curve is expressed by a function that is symmetrical with respect to a time axis at the center of the power generation time.

[0010] By using such a function, it is possible to reproduce the clear weather power generation curve with fewer parameters.

[0011] In the second step of the method for predicting power generation according to the present invention, the extracted date has a total sunny time period of 40% or more of the total power generation time period.

[0012] By configuring in this way, the number of days on which the clear weather power generation curve can be calculated increases, which increases the number of data points for constructing the annual transition curve and improves the accuracy of the prediction.

[0013] In addition, in the method for predicting power generation according to the present invention, the weather coefficients are defined for each hour or month.

[0014] This configuration allows for more precise prediction of power generation.

[0015] Furthermore, the method for predicting power generation according to the present invention is characterized in that in step 2, the power generation prediction device calculates the clear weather power generation curve based on the power generation value at the time when the power generation value is below the rated output of the power conditioner, and in step 7, the minimum value of the predicted power generation and the rated output calculated for each hour from the clear weather power generation model and the weather coefficient is adopted as the predicted power generation at that hour.

[0016] With this configuration, it becomes possible to predict the power generation even in the case where, for example, a peak cut occurs in a photovoltaic power generation system with an overload of solar cells.

[0017] In addition, the method for predicting power generation according to the present invention is characterized in that the predicted power generation after the target prediction date is corrected by comparing the power generation output from the power conditioner actually measured on the target prediction date with the predicted power generation.

[0018] With this configuration, it is possible to further improve the accuracy of prediction of generated power.

[0019] A photovoltaic power generation system according to the present invention is a photovoltaic power generation system including a solar cell, a power conditioner, and a power generation power prediction device, wherein the power generation power prediction device executes the prediction method according to any one of claims 1 to 6.

[0020] Such a photovoltaic power generation system makes it possible to predict the power generated by sunlight without the need for a solar radiation meter.

[0021] The power generation prediction device according to the present invention is a power generation prediction device for predicting the power generation of a photovoltaic power generation system equipped with a solar cell and a power conditioner, and is equipped with an arithmetic processing unit and a storage device, wherein the arithmetic processing unit executes the prediction method according to any one of claims 1 to 6.

[0022] The power generation prediction device can be incorporated into a photovoltaic power generation system that includes an existing solar cell and a power conditioner, and as a result, a photovoltaic power generation system that can predict power generation can be constructed.

[0023] According to the present invention, it is possible to obtain a method for predicting photovoltaic power generation, a photovoltaic power generation prediction device, and a photovoltaic power generation system that are capable of predicting power generation at a desired date and time without requiring an actinometer.

[0024] FIG. 1 is a conceptual diagram showing the main components of a solar power generation system 100. FIG. 2(A) is a graph illustrating the time transition (time dependency) of the actual measured values ​​of power generation on a clear day all day long. FIG. 2(B) is a graph illustrating an approximation curve (referred to as a clear-weather power generation curve) that reproduces the time dependency of the power generation (measured values). FIG. 3 shows the time dependency of the power generation of the solar cell 2, showing an example of a partially clear day. FIG. 4 is a graph showing the annual transition of the total power generation amount on clear days, and an annual transition curve that approximates the annual transition of the total power generation amount on clear days. FIG. 5(A) shows an example of a partially clear day that cannot be corrected to a clear-weather day by interpolation, and FIG. 5(B) shows an example of a clear-weather power generation curve created based on the actual measured values ​​of power generation on a partially clear day.

[0025] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the following embodiments are not intended to limit the scope of the present invention. Furthermore, the same or similar components will be designated by the same reference numerals, and their description may be omitted.

[0026] The term "generated power" used below refers to the power generation capacity per unit time, expressed, for example, in units of kW, and "power generation amount" refers to the integrated value of generated power over a specified period of time (for example, the total power generation time in one day), expressed, for example, in units of kWh.

[0027] (Embodiment 1) FIG. 1 is a conceptual diagram showing the main components of a solar power generation system 100. The solar power generation system 100 of embodiment 1 includes a power generation power prediction device 1, a solar cell 2, and a power conditioner 3. The direct current (DC) power generated by the solar cell 2 is converted to alternating current (AC) power via the power conditioner 3 (referred to as PCS) and output to a load (not shown). A known power conditioner can be used for the solar cell 2 and the PCS 3. The PCS 3 can receive power generated by the solar cell 2, control the power generation efficiency of the solar cell 2 according to the IV characteristics of the solar cell 2, and output the generated power. For example, the PCS 3 can be a device that controls the output of the solar cell 2 to maximize the power generation efficiency of the solar cell 2 using a hill-climbing method or the like.

[0028] The PCS 3 outputs the power generation value (output value of the power generation) to the power generation prediction device 1. Note that a power meter that measures the power generation output from the PCS 3 may be provided separately, and the power meter may output the power to the power generation prediction device 1. Therefore, the power generation prediction device 1 can obtain the power generation of the solar cell 2 that is output via the PCS 3.

[0029] The power generation prediction device 1 includes a processing unit 11 and a storage device 12. The processing unit 11 of the power generation prediction device 1 receives the power generation value that changes from moment to moment from the PCS 3, records it together with the date and time in the storage device 12, and is also capable of reading out the power generation value recorded in the storage device 12. The processing unit 11 can use a known microprocessor or the like, and the storage device 12 can use a known device that stores data using a semiconductor memory or an electromagnetic method.

[0030] The power generation prediction device 1 may be installed at the location where the solar cell 2 is installed, or may be located in a different location, such as a remote location connected to a network. The power generation prediction device 1 may be connected to the PCS 3 or a power meter via a data communication means such as the Internet or a LAN and receive power generation values. The power generation prediction device 1 may be located, for example, in a monitoring room or a remote terminal. The power generation prediction device 1 may also include a processing unit 11 and a storage device 12. Therefore, the power generation prediction device 1 may be configured as a single device using an independent computer such as a so-called IoT device. Alternatively, the power generation prediction device 1 may use a cloud-based computer, a processing unit and storage device incorporated in a monitoring device that monitors or visualizes power generation, or a processing unit and storage device included in a control device of the PCS 3. Furthermore, each step of the prediction method described below may be performed by multiple different computers or other devices. The power generation prediction device 1 may be configured with multiple processing units 11 and / or storage devices 12.

[0031] The power generation prediction device 1 can be connected to the Internet (line) 4 via a network interface (not shown), and can obtain weather information such as cloud cover, temperature, humidity, and weather forecasts via the Internet 4. The weather information and weather forecasts are recorded in a storage device 12 together with the date and time, and the weather information and weather forecasts recorded in the storage device 12 can be read out. The power generation prediction device 1 can obtain current or past weather information as weather information.

[0032] The power generation prediction device 1 can output a video signal and may also include a display device 13, such as a flat panel display. In this case, the power generation prediction device 1 can output the power generation value of the solar cell 2 to the display device 13 and visualize it.

[0033] The power generation prediction device 1 can receive instructions from an operator via an input device 14 such as a keyboard, a touch panel, or a voice input. The display device 13 can also be configured as the input device 14 by being equipped with a touch panel function.

[0034] The storage device 12 stores the power generation value of the solar cell 2 for each day in the past year at predetermined time intervals. For example, the storage device 12 stores the power generation value for the past year (January 1 to December 31) at intervals of, for example, one minute.

[0035] A method for predicting the power generated by the solar cell 2 using the power generation prediction device 1 will be described below.

[0036] Step 1: Prepare power generation and weather data before the prediction date. Prepare power generation data and weather data for January 1st to December 31st, before the prediction date (target prediction date). The power generation prediction device 1 stores and accumulates historical data in a database in which the hourly power generation data of the solar cell 2 output via the PCS 3 for each month, day, hour, and minute, and weather information (cloud cover, temperature, humidity) including at least cloud cover obtained from the Internet 4, etc., are correlated with each other on a daily basis as power generation data and weather data. Note that power generation data and weather data from at least the past year prior to the day (current day) on which the prediction is performed are sufficient, and the starting point of the power generation data and weather data is not limited to January 1st. Furthermore, historical power generation data and weather data from more than one year may be stored in the storage device 12.

[0037] Step 2: Clear days are extracted (selected) from the data of past power generation, and a clear weather power generation curve (clear weather power generation curve) is obtained from the actual measured values ​​of power generation on those clear days.

[0038] A day with clear skies from the power generation start time (or sunrise) to the power generation end time (or sunset) is extracted from the actual measured values ​​of power generation in the past stored in the storage device 12. Hereinafter, the extracted (selected) day may be referred to as an extracted day. Note that the period from the power generation start time (or sunrise) to the power generation end time (or sunset) may be simply referred to as an entire day, and a day with clear skies from the power generation start time to the power generation end time may be referred to as an all-day clear day.

[0039] Figure 2(A) is a graph illustrating the time transition (time dependency) of the actual measured value of power generation on a clear day all day long. The vertical axis represents the value of the power generation normalized by the rated power of PCS3 (= [power generation value] / [rated power of PCS3]), and the horizontal axis represents time. Figure 2(B) is a graph illustrating an approximation curve (referred to as a clear weather power generation curve) that reproduces the time dependency of the actual measured value of power generation. Hereinafter, the generated power and power generation amount will be values ​​normalized by the rated power of PCS3. The rated power of PCS3 is the maximum power value that PCS3 can output.

[0040] It is possible to extract sunny days all day long from the shape of a graph showing the time dependency of the power generation (actual measured values). For example, as shown in FIG. 2A, the shape of the graph showing the time dependency of the power generation (actual measured values) is similar to a Gaussian distribution overall, and is a gentle curve, so it can be determined that the data is of power generation on sunny days all day long. As will be described later, it is possible to automatically determine whether the weather is sunny at each time from the value of power generation, and days determined to be sunny all day long can be extracted as sunny days all day long. Such sunny days can be extracted by the calculation processing device 11 sequentially reading past measured values ​​of power generation stored in the storage device 12 and automatically determining whether they are sunny days all day long. Days determined to be sunny all day long are registered in the storage device 12 as sunny days all day long.

[0041] Furthermore, the clear days to be extracted may not only be clear days all day long, but also days that are partially clear at a specific time (referred to as partially clear days). Figure 3 shows the time dependency of the power generated by the solar cell 2, illustrating an example of a partially clear day. In Figure 3, it can be seen that the power generated fluctuates significantly at the times indicated by the white arrows. The power generated at the times indicated by the white arrows indicates values ​​that deviate from the power generated on a clear day based on the time dependency of the power generated on a clear day. The power generated at these times can be linearly interpolated using the power generated before and after the time determined to represent the power generated on a clear day, and corrected to the power generated on a clear day. Partially clear days can be corrected to clear days all day long by interpolation.

[0042] A partially clear day that can be corrected to an all-day clear day by interpolation can be registered as an interpolatable partially clear day in the storage device 12. Alternatively, for a partially clear day that can be corrected to an all-day clear day by interpolation, the deviating value of the power generation data may be corrected by interpolation, and the corrected power generation data may be registered in the storage device 12 as an all-day clear day.

[0043] The time dependence of the power generation (measured value) on the extracted clear day (or partial clear day that can be corrected to clear day) is approximated by a predetermined continuous function (F(t): t is time) to obtain a power generation curve for a clear day (referred to as a clear day power generation curve). The obtained clear day power generation curve shows the time dependence of the power generation on a clear day. The continuous function (F(t)) that reproduces the power generation can be obtained from the time dependence of the power generation (measured value) by regression analysis such as the least squares method.

[0044] As a predetermined function, for example, as shown in Figure 2 (B), the time is divided into three regions, A, B, and C, in order of progression of time, with the start time of region A being the power generation start time and the end time of region C being the power generation end time, and the approximate function of region C being the inverse (axisymmetric) of the function of region A. The approximate functions are continuously connected at the boundaries between regions A and B and between regions B and C. By including information on the start and end times of power generation and by making the functions of regions A and C axisymmetric to each other, the number of parameters determining the approximate function can be reduced. In other words, it is possible to reproduce the clear weather power generation curve with fewer parameters.

[0045] As a specific function, for example, a combination of trigonometric functions can be used in regions A and C, and a quadratic function can be used in region B. The power generation start time (the time when the generated power begins to exceed 0) is Ts, the power generation end time (the time when the generated power decreases to 0) is Te, and the time in regions A and C is γ. If time is t, the range for region A is Ts≦t≦Ts+γ, and the range for region C is Te-γ≦t≦Te. The following functions can be used as approximate formulas for regions A and C. Region A: dF(t) / dt=α*sin(2π(t-Ts) / β) ... (Equation 1) Region C: dF(t) / dt=α*sin(2π(Te-t) / β) ... (Equation 2) Here, dF(t) / dt is the first-order derivative of F(t) with respect to time.

[0046] In region B, the approximation function is a quadratic function of time t, with the coefficient of the quadratic term of t being negative, and smoothly continuing with the approximation formula for region A at the boundary between region A and region B (t = Ts + γ), and smoothly continuing with the approximation formula for region C at the boundary between region B and region C (t = Te - γ). Note that "smoothly continuing" means being continuous up to at least the first-order differential. The quadratic function that determines the approximation formula for region B is maximized at the center of the time (t = (Ts + Te) / 2) when solar cell 2 is generating electricity, and is uniquely determined by parameters α, β, and γ.

[0047] The power generation start time and power generation end time can be set to the sunrise and sunset times determined from the date, latitude, and longitude, but can also be modified taking into account the influence of the topography and objects around the solar cell 2.

[0048] Since the clear weather power generation curve is obtained based on the actual measured value of the power generation, the continuous function that reproduces the clear weather power generation curve will include the influence of the IV characteristics of the solar cell 2, the performance of the PCS 3 that controls the output of the solar cell 2, etc.

[0049] The function form used to approximate the clear-weather power generation curve is not limited to the above and can be set as appropriate. However, using a symmetric function as described above allows the clear-weather power generation curve to be reproduced using fewer parameters, making it easier to analyze the trend of the clear-weather power generation curve over time. Furthermore, if the clear-weather power generation curve obtained from power generation data exhibits a complex shape, it is possible to obtain an approximation formula using a neural network based on the universality theorem.

[0050] Step 3: Obtain an annual transition curve showing the annual change in the total daily power generation on clear days. The power generation at each time point is integrated from the clear-day power generation curves obtained in Step 2 for multiple days, and the total daily power generation for each day (referred to as the clear-day total power generation) is calculated to obtain the annual change in the clear-day total power generation (referred to as the annual transition curve). Figure 4 is a graph showing the annual change (day dependency) in the clear-day total power generation, and further shows an annual transition curve approximating the annual change in the clear-day total power generation. The vertical axis represents power generation, and the horizontal axis represents time (days). Each point in Figure 4 represents the clear-day total power generation, and the solid line represents the annual transition curve. Figure 4 is not limited to clear-days throughout the day, but also includes the clear-day total power generation calculated using the clear-day power generation curve for partial clear days. The clear-day total power generation is the value obtained by integrating F(t) from the power generation start time Ts to the power generation end time Te on the date (τ) of each point. Note that F(t) also depends on the date (τ) and is a function of τ and t, so strictly speaking, F(t) = F(τ, t).

[0051] An approximation function (annual transition approximation function G(τ)) that reproduces the day dependency of the total power generation amount on clear days shown at each point in Figure 4 is obtained by regression analysis. Here, τ is the number of days. The annual transition curve can be expressed by the function G(τ). The function G(τ) may be composed of, for example, a combination of two periodic functions.

[0052] As a specific function, for example, an annual transition curve may be constructed by linearly combining a periodic function with a cycle of one year and a periodic function with a cycle of six months, as shown in Equation 3: G(τ) = a * sin(2π * (τ - b) / 365) + c * sin(2π * (τ - d) / 182.5) + e ... (Equation 3), where a, b, c, d, and e are parameters. By approximating the annual transition curve with a function constructed by linearly combining functions with two different cycles, as in Equation 3, it is possible to reflect the influence of the annual transition of the sun's orbit, as well as climatic and geological factors.

[0053] Step 4: The clear-weather power generation curve is scaled to fit the annual transition curve to obtain a clear-weather power generation model (Fm(t)). Because each clear-weather power generation curve is calculated for each clear-weather day, there may be variations in the relationship between the clear-weather total power generation calculated from each clear-weather power generation curve. Because the clear-weather total power generation changes smoothly, each clear-weather power generation curve is multiplied by a correction coefficient to scale it to fit the annual transition curve. For example, if the clear-weather total power generation calculated using the clear-weather power generation curve for the clear-weather day extracted in step 2 (referred to as the extracted day) does not match the total power generation calculated from the annual transition curve (referred to as the calculated total power generation), the clear-weather power generation curve is multiplied by a correction coefficient Cm to obtain a clear-weather power generation model. The correction coefficient Cm for the extracted day (τ) is calculated as follows: Cm(τ) = [calculated total power generation] / [clear day total power generation] = G(τ) / ∫F(τ, t) ... (Equation 4) Note that Cm is a constant in the sense that it is independent of the time (t), but is determined for each extracted day (τ) (Cm = Cm(τ)). Therefore, Cm(τ) is a function of the day (τ). The clear day power generation model (Fm(τ, t)) can be obtained by multiplying the clear day power generation curve for the extracted day (τ) by the calculated correction coefficient Cm. Fm(τ, t) = Cm(τ) x F(τ, t) ... (Equation 5)

[0054] Step 5: Obtain all clear-sky power generation models for the year. To obtain clear-sky power generation models for days other than the extraction date, the clear-sky power generation models for days other than the extraction date are generated by interpolating the clear-sky power generation models (i.e., Fm(τ1, t) and Fm(τ2, t)) for the extraction date (τ1; where τ1 < τ) and the extraction date (τ2; where τ2 > τ) immediately before and after the day for which the clear-sky power generation model is to be generated (the target date for generating the clear-sky power generation model (τ)) in the past data. For example, linear interpolation is used as the interpolation method. If the interpolated clear-sky power generation model is designated as Fm'(τ, t), and linear interpolation is used as an example, Fm'(τ, t) is expressed as follows: Fm'(τ,t) = Fm(τ1,t) + {(τ - τ1) / (τ2 - τ1)} × {Fm(τ2,t) - Fm(τ1,t)} (Equation 6) Note that a correction coefficient Cm'(τ) may be calculated and Fm'(τ,t) may be scaled so that the clear weather power generation model Fm'(τ,t) obtained by interpolation for the target day (τ) for generating the clear weather power generation model matches the annual trend curve. The correction coefficient Cm'(τ) can be calculated using the following equation, as in step 4. Cm'(τ) = [calculated total power generation] / [total power generation of interpolated clear weather power generation model] = G(τ) / ∫Fm'(τ,t) (Equation 7) The clear weather power generation model Fm(τ,t) for the target day (τ) for generating the clear weather power generation model can be calculated as follows. Fm(τ, t) = Cm'(τ) × Fm'(τ, t) (Equation 8) The interpolated clear weather power generation model is expressed as Fm'(τ, t) to distinguish it from Fm(τ, t) for ease of understanding.

[0055] As a result, a clear-weather power generation model for all days in a year can be obtained. All generated clear-weather power generation models are saved in the storage device 12 together with the date and time. The clear-weather power generation model shows the time dependency of power generation that reflects annual trends based on the actual measured values ​​of power generation output from the power conditioner 3 on clear-weather days. Therefore, it is possible to comprehensively reflect not only annual changes in solar radiation, but also the influence of climate such as temperature, and the characteristics of the solar cell 2 and the power conditioner 3.

[0056] In addition, if the past data does not contain any extraction dates before or after the day for which the clear-weather power generation model is to be generated (the target day for generating the clear-weather power generation model (τ)), and only contains one of the extraction dates (τ1 or τ2) before or after the target day for generating the clear-weather power generation model (τ), the clear-weather power generation model for the most recent extraction date before or after that day is used as the clear-weather power generation model for that target day for generating the clear-weather power generation model. However, in this case, the most recent clear-weather power generation model may be calculated by multiplying the correction coefficient Cm' calculated using the most recent clear-weather power generation model to fit the annual transition curve. Cm' = [calculated total power generation] / [total power generation of the most recent clear-weather power generation model] (Equation 9) For example, if one of τ1 or τ2 is τk, then: Cm'(τ) = G(τ) / ∫Fm(τk, t) (Equation 10) The clear-weather power generation model Fm(τ, t) for the target day for generating the clear-weather power generation model (τ) can be calculated as follows: Fm (τ, t) = Cm' (τ) × Fm (τk, t) ... (Formula 11)

[0057] Furthermore, to reproduce the power generation start time and power generation end time of the target day for generating the clear weather power generation model, the power generation start time and power generation end time may be calculated from the latitude and longitude, and the time axis may be scaled (the power generation time may be reduced or expanded) for the clear weather power generation model obtained by multiplying the calculated time by the constant. For example, if the power generation start time and power generation end time calculated from sunrise and sunset on the target day for generating the clear weather power generation model are T1i and T1e, respectively, and the power generation start time and power generation end time of the most recent extracted day are T0i and T0e, respectively, the power generation time of the clear weather power generation model for the extracted day may be scaled (multiplied by (T0e-T0i) / (T1e-T1i)) around the midpoint between the power generation start time and the power generation end time. For example, if the midpoint between the power generation start time and the power generation end time is Tc = (T1e + T1s) / 2, correction can be performed using the following equation 12. Fm(τ, t) = Cm'(τ) x Fm(τk, Tc + α x (t - Tc)) ... (Equation 12) Note that the clear weather power generation model Fm(τ, t) calculated by the above interpolation may also be time-scaled (or adjusted) as shown below to reproduce the power generation start and end times on the target day for generating the clear weather power generation model. Fm(τ, t) = Cm'(τ) x Fm'(τ, Tc + α x (t - Tc)) ... (Equation 13)

[0058] Step 6: Calculate a weather coefficient for each weather condition. The measured power generation amount for each weather condition (cloud cover) is compared with the clear-weather power generation model. The average value of the ratio between the measured power generation amount for each cloud cover and the power generation amount of the clear-weather power generation model at each time is defined as the weather coefficient (Rw), and the weather coefficient for each cloud cover is calculated. The power generation prediction device 1 stores the weather coefficient in the storage device 12 in association with the cloud cover. Rw (cloud cover) = <[Measured power generation amount] / [Power generation amount of the clear-weather power generation model]> (Equation 14) where < > indicates the average value. The average is, for example, the arithmetic mean of multiple ratios obtained at each time. Note that, as mentioned above, the cloud cover for the corresponding time can be obtained from the Internet 4, and weather data previously stored in the storage device 12 can be used. The weather coefficient is determined for each cloud cover, making it possible to calculate the power generation amount corresponding to the cloud cover from the power generation amount of the clear-weather power generation model.

[0059] As shown below, the cloud cover can be set to, for example, five levels: Level 1: Cloud cover 10% or less (equivalent to clear skies) Level 2: Cloud cover 11% to 25% or less Level 3: Cloud cover 26% to 50% or less Level 4: Cloud cover 51% to 84% or less Level 5: Cloud cover 85% to 100% or less The weather coefficient Rw is determined as a function Rw(L) that depends on the level L. Note that Rw(L) can be calculated from the actual measured values ​​of power generation for each date stored in the storage device 12, and is therefore calculated as the average value of multiple ratios obtained according to the cloud cover L. Rw(L) = <[Actual power generation] / [Power generation by the clear skies power generation model]> (Equation 15) Here, < > indicates the average value.

[0060] Step 7: Predict the power generation for the target prediction date. The power generation prediction device 1 obtains the weather forecast for the target prediction date for the area where the solar cell 2 is installed from the Internet 4. The weather forecast includes a forecast of cloud cover (L) for each time (t) on the target prediction date (τ). The power generation prediction device 1 reads the clear sky power generation model (Fm(τ, t)) and weather coefficient (Rw(L)) corresponding to the date (τ) of the target prediction date from the storage device 12, and multiplies the power generation of the clear sky power generation model by the weather coefficient corresponding to the cloud cover for each time (t) to calculate the power generation, thereby obtaining the predicted power generation (Pf(τ, t)), which is the predicted value of the power generation. As shown below, the predicted power generation (Pf(τ, t)) can predict the power generation for each time (t) and provides the time dependency of the predicted power generation. Pf(τ, t) = Rw(L) x Fm(τ, t) ... (Equation 16) By the above steps 1 to 7, the power generation prediction device 1 can obtain the predicted power generation Pf(τ, t) for each time (t) on the prediction target day (τ).

[0061] A program for executing steps 1 to 7 in this order is stored in the storage device 12, and the arithmetic processing device 11 predicts the power generation according to the stored program. The arithmetic processing device 11 is capable of outputting the power generation for the prediction target day to the display device 13, and may also be configured to output the power generation to a control device such as an external computer. The form of the storage device 12 is not particularly limited, and it may be built into the arithmetic processing device 11 or provided externally.

[0062] Furthermore, the power generation prediction device 1 that executes the above steps can take various forms as described above.

[0063] It is possible to automatically execute all of the processes from step 1 to step 7 above, but an operator may intervene in some or each of the steps above and execute steps 1 to 7 in order while checking the results, etc. This does not mean that human operation is excluded from the execution of the processes.

[0064] The power generation prediction device 1 can be incorporated into an existing photovoltaic power generation system, and therefore, a photovoltaic power generation system 100 capable of predicting photovoltaic power generation can be easily constructed.

[0065] (Embodiment 2) In step 2 above, the extracted dates (clear days to be extracted) were all-day clear days as well as partially clear days that can be corrected to all-day clear days. However, partially clear days that cannot be corrected to all-day clear days by interpolation can also be used as extracted dates. As will be explained below, the power generation power prediction device 1 can extract days for which a clear weather curve can be calculated from the power generation power data stored in the storage device 12, and calculate the clear weather curve.

[0066] Figure 5(A) shows an example of a partially clear day that cannot be corrected to a clear day all day by interpolation, and Figure 5(B) shows an example of a clear day power generation curve created based on the actual measured values ​​of power generation on a partially clear day. The clear day power generation curve F(t) shown in Figure 5(B) combines trigonometric functions and quadratic functions as described above, and uses a function that is symmetrical with respect to the central time axis of the power generation time.

[0067] In Figure 5(A), the power generation values ​​for time period A and time period B can be considered to be those for clear skies. However, the power generation values ​​for other time periods fluctuate significantly in increments of about five minutes, making it impossible to obtain the power generation on a clear skies day by interpolating the power generation values. However, if clear skies (clear skies) account for a certain percentage of the total power generation time (the time from the start time of power generation to the end time of power generation), for example, 40% or more, it is possible to obtain a function that reproduces the clear skies power generation curve using only the measured power generation values ​​for time periods that can be considered clear skies. Since the number of days for which the clear skies power generation curve can be calculated increases, the number of data points for constructing the annual transition curve also increases, resulting in improved prediction accuracy.

[0068] In the example shown in Figure 5(A), the power generation start time is 6:46 and the power generation end time is 17:23, for a total power generation time of 637 minutes. Time periods A and B that can be considered clear are 126 minutes and 148 minutes, respectively, for a total of 274 minutes, making the total clear weather time 43% of the total power generation time. Figure 5(B) shows the results of a regression analysis performed using only the power generation values ​​for time periods A and B, excluding power generation values ​​outside of time periods A and B. As shown in Figure 5(B), even on a partially clear day that cannot be corrected to a clear day all day by interpolation, a clear weather power generation curve can be obtained as long as it is a day on which a clear weather curve can be calculated.

[0069] In addition, days when the weather is clear all day, days when power can be obtained that can be corrected to a clear day by interpolation, and days when it is difficult to correct to a clear day all day but when power generation equivalent to that of a clear day can be obtained for a specified period of time or more are sometimes referred to as days when the clear day curve can be calculated.

[0070] The arithmetic processing unit 11 of the power generation forecasting device 1 can automatically determine whether or not each day stored in the storage device 12 is a day on which the clear weather curve can be calculated, based on the power generation stored for each day in the storage device 12. A method for automatically determining whether or not a day on which the clear weather curve can be calculated is possible will be described below.

[0071] In order to automatically determine days on which the clear weather curve can be calculated, it is necessary to determine from the generated power data shown in Figure 5(A) that time periods in which the generated power fluctuates significantly on a minute-by-minute basis are not clear weather and to exclude them. The calculation processing unit 11 reads the generated power data stored in the storage device 12, determines a predetermined time interval ΔT (e.g., 5 minutes), and determines whether each time is clear weather or not based on its dependency on the predetermined time interval. The calculation processing unit 11 can execute the following determination process for each time t in order, starting from the start time Ts.

[0072] Judgment process 1: If the total fluctuation in power generation from time t to time t + ΔT (|P(t) - P(t + ∂t)| + |P(t + ∂t) - P(t + 2∂t)| + ... + |P(t + ΔT - ∂t) - P(t + ΔT)|) is equal to or greater than a certain value, it is judged (determined) that the day is not sunny. Note that P(t) is the power generation at time t, ∂t is the measurement time interval (or the output time interval of the actual measured value) of the power generation P(t), and || indicates the absolute value. Judgment process 2: If the integrated value of power generation from time t to time t + ΔT is smaller than the integrated value of power generation for both adjacent ΔT, that is, smaller than the integrated value of power generation from time t - ΔT to time t and smaller than the integrated value of power generation from time t + ΔT to time t + 2ΔT, it is judged (determined) that the day is not sunny. Determination process 3: If the integrated value of the power generation from time t to time t+ΔT is less than the average of the integrated values ​​of the power generation from time t to time t+ΔT for 15 days before and after, it is determined that the day is not sunny. The calculation processing device 11 executes the above determination processes 1, 2, and 3, and if it is determined that any one of them is not sunny, it determines (judges) that time t is not sunny. Note that the order in which determination processes 1, 2, and 3 are executed is arbitrary and is not limited.

[0073] The arithmetic processing device 11 excludes the times determined not to be clear skies by the above-mentioned judgment process from the power generation time (from the power generation start time to the power generation end time), and determines the remaining time as clear skies. The arithmetic processing device 11 calculates the total amount of time determined to be clear skies. The arithmetic processing device 11 recognizes a day when the total amount of time determined to be clear skies is equal to or greater than a predetermined percentage (e.g., 40%) of the power generation time as a day on which a clear skies curve can be calculated, and registers this in the storage device 12 as a day on which a clear skies curve can be calculated, and can also register the time determined to be clear skies on that day on which a clear skies curve can be calculated.

[0074] The calculation processing unit 11 of the power generation prediction device 1 can perform regression analysis using only the actual measured values ​​of power generation during times determined to be sunny on days when the sunny weather curve can be calculated, which are stored in the memory device 12, to obtain a sunny weather power generation curve.

[0075] (Embodiment 3) In the above embodiment, the weather coefficients are determined depending on the amount of cloud cover. The weather coefficients may also be defined to vary depending on the season or time of day. For example, the amount of solar radiation irradiating the solar cell 2 is considered to be affected not only by the amount of cloud cover but also by the amount of water vapor in the atmosphere, and is also considered to be affected by temperature and humidity. Therefore, the weather coefficients may be calculated for each month depending on the weather and recorded in the storage device 12 of the power generation prediction device 1. A weather coefficient may be selected depending on the month of the target prediction date, and the power generation may be predicted in conjunction with the cloud cover in the weather forecast.

[0076] Furthermore, the weather coefficients may also depend on the time of day, as the ratio of the component of sunlight that is directly incident (direct component) to the component of sunlight that is incident after being scattered by clouds (scattered component) changes depending on the orientation of the solar cell 2. For this reason, the weather coefficients may be calculated for each hour or each time zone and recorded in the storage device 12 of the power generation prediction device 1, and the power generation may be predicted by selecting a weather coefficient according to the time of the day to be predicted and combining it with the cloud cover in the weather forecast.

[0077] When the weather coefficient (Rw) is defined for each month, the weather coefficient (Rw) can be calculated for each month according to Equation 14 or Equation 15 to determine the average value. Since the weather coefficient (Rw) depends on the month (M) and cloud cover (L), it can be expressed as follows: Rw(L) = Rw(M, L) ... (Equation 17). Furthermore, Rw(M, L) may be set for each of several months, for example, every three months, or for each season. As shown in FIG. 4, the annual transition has two maxima and two minima, and therefore seasons may be set to include these maxima and minima.

[0078] Furthermore, when the weather coefficient (Rw) is defined for each time period, for example, the power generation time (from the start time to the end time of power generation) can be divided into multiple time segments, for example, three segments, and determined. The power generation time can be divided, for example, equally. Furthermore, for example, the continuous function F(t) that reproduces the power generation may be composed of three types of functions as in the above example and divided into three regions, A, B, and C, as shown in FIG. 2(B). When the power generation time is defined for each time segment, the weather coefficient (Rw) can be calculated for each time segment according to Equation 14 or Equation 15, and the average value can be determined. Because the weather coefficient (Rw) depends on the time segment (J) and the cloud cover (L), it can be expressed as Rw(L) = Rw(J, L) (Equation 18).

[0079] When the weather coefficient (Rw) is defined for each month and each time segment, the weather coefficient (Rw) can be calculated by averaging the values ​​obtained for each month (M) and each time segment (J) according to Equation 15. In this case, the weather coefficient (Rw) depends on the month (M) and the time segment (J) in addition to the cloud cover (L), and can therefore be expressed as follows: Rw(L) = Rw(M, J, L) (Equation 19) In this way, by defining the weather coefficient (Rw) as a function that depends on the month, season, or time, it becomes possible to predict the power generation more precisely.

[0080] (Fourth Embodiment) In the photovoltaic power generation system 100, the solar cells 2 may be overloaded to increase the total amount of power generation. In such a configuration, for example, around noon when the amount of solar radiation is high, the power generation that can be output from the solar cells 2 may exceed the rated capacity of the PCS 3. As shown in the above example, the power generation output from the solar cells 2 may be limited by the rated power of the PCS 3. Note that a state in which the power generation is limited by the rated power of the PCS 3 and no more power is output is referred to as peak shaving. When peak shaving occurs, the power generation output from the PCS 3 becomes equal to the rated power of the PCS 3. The following describes power generation prediction of the photovoltaic power generation system 100 when peak shaving occurs. In this embodiment, the following steps are modified from the power generation prediction method of the first embodiment.

[0081] Step 2: In the above step 2, the extracted power generation data on clear days is checked for peak shaving at predetermined time intervals, for example, every minute. Times when peak shaving occurs are excluded from the clear days, and a clear day power generation curve is calculated from times when peak shaving does not occur. Whether peak shaving has occurred is determined by determining whether the value of the power generation output from PCS 3 is the same as the rated output (rated power) of PCS 3. An approximation function that reproduces the power generation at the remaining times that have not been excluded, i.e., times when the value of the power generation output from PCS 3 is equal to or less than the rated output of PCS 3, is found by regression analysis such as the least squares method, and the clear day power generation curve is obtained.

[0082] Note that, for example, due to the influence of the control characteristics of the PCS 3, when peak shaving occurs, the generated power may fluctuate at a value lower than the rated output. In this case, the value of the generated power output from the PCS 3 is not necessarily exactly the same as the rated output of the PCS 3. It may be determined that peak shaving has occurred when the value of the generated power output from the PCS 3 is near the rated output, i.e., the difference from the rated output is within a predetermined range (e.g., within 5% of the rated output) and remains within the predetermined range for a predetermined period of time or more (e.g., 10 minutes or more). Therefore, it is determined whether the value of the generated power output from the PCS 3 is substantially equivalent to the rated output of the PCS 3. Substantially equivalent means that the values ​​are the same or the difference is within a predetermined range.

[0083] Step 7: In the process of Step 7, the power generation prediction device 1 compares the predicted power generation Pf(τ, t) predicted for each time (t) on the prediction target day (τ) with the rated output of the PCS 3, and adopts the minimum value as the predicted power generation, which is the predicted value of the power generation at that time. Pf(τ, t) = min(Pf(τ, t), [rated output of PCS 3]) (Equation 20) In Equation 20, "min" is a function that outputs the minimum value. If Pf(τ, t) and the rated output of PCS 3 are the same value, that value is output.

[0084] (Embodiment 5) By comparing the power generation amount, which is the actual power generation amount measured on the target prediction date or the integrated value of the actual power generation amount, with the predicted power generation amount or the integrated value of the predicted power generation amount, it is possible to correct the predicted power generation amount after the target prediction date, thereby further improving the accuracy of the power generation prediction.

[0085] Correction method 1: On the target prediction date, the actually measured power generation amount and the predicted power generation amount are compared, and the ratio is calculated as the power generation amount correction ratio (Rp). Rp = [actually measured power generation amount] / [predicted power generation amount] (Equation 21) The predicted power generation amount is multiplied by the obtained power generation amount correction ratio to correct the predicted power generation amount thereafter (on the target prediction date). For example, the power generation amount correction ratio is calculated from the power generation amount obtained by integrating the actually measured power generation amount and the predicted power generation amount on the target prediction date, X month Y day, and the predicted power generation amount is corrected by multiplying the predicted power generation amount by the power generation amount correction ratio. The predicted power generation amount on and after X month Y+1 day is corrected. The power generation amount correction ratio may be stored in the storage device 12.

[0086] Furthermore, if the target forecast date, X month Y day, is a day on which the clear weather curve can be calculated, the clear weather power generation curve may be calculated from the data of the actually measured power generation, and the integrated value of the clear weather power generation curve may be used in place of the [actually measured power generation amount] in Equation 21, and the integrated value of the predicted clear weather power generation model for X month Y day may be used in place of the [predicted power generation amount]. The obtained Rp may be used to correct the clear weather power generation model for the forecast target date and thereafter, and the corrected clear weather power generation model may be stored in the storage device 12. Note that since the clear weather model can be corrected using Rp, it is referred to as the clear weather model correction ratio.

[0087] Correction Method 2: On the target prediction date, it is also possible to further adjust the power generation after the start of power generation. That is, the power generation predicted for each hour before the target prediction date may be corrected using the actual measured power generation value obtained after the start of power generation. If the weather is clear at a predetermined time after the start of power generation on the target prediction date, the integrated value of the actually measured power generation during that clear time is compared with the integrated value of the predicted power generation. The ratio (= [integrated value of actually measured power generation] / [integrated value of predicted power generation]) is calculated as the power generation correction ratio Rp, and the predicted power generation from the predetermined time onward is corrected by multiplying the predicted power generation by the correction ratio. For example, if the start time of power generation is time A and the weather is clear from time A to time B, the power generation correction ratio Rp is calculated from the integrated value obtained by integrating the actual measured power generation values ​​from time A to time B and the integrated value obtained by integrating the predicted power generation. The predicted power generation from time B onward is corrected by multiplying the calculated power generation correction ratio Rp by the calculated power generation correction ratio Rp.

[0088] Correction Method 3: It is also possible to correct the weather coefficient. At each time on the target forecast day, the actual measured value of the power generation and weather observation data, such as cloud cover, can be obtained via the Internet. Therefore, it is possible to calculate the weather coefficient Rw using Equation 14 and update it as needed. The calculated weather coefficient Rw may also be updated as a weather coefficient Rw defined for each month and each time segment.

[0089] According to the present invention, it is possible to provide a power generation system that can predict the power generation amount of a solar power generation system from past power generation data, weather information, and a weather forecast for the predicted date, without using a solar radiation meter. Existing solar power generation systems can be utilized, and the present invention has high industrial applicability.

[0090] REFERENCE SIGNS LIST 100 Photovoltaic power generation system 1 Power generation prediction device 11 Processing device 12 Storage device 13 Display device 14 Input device 2 Solar cell 3 Power conditioner 4 Internet

Claims

1. A method for predicting power generation in a solar power generation system including a solar cell and a power conditioner, a first step of preparing historical data that associates power generation data of the solar cell output via the power conditioner with weather data for at least one year prior to a target date for power generation prediction; a second step of obtaining a clear weather power generation curve from the power generation values ​​of the selected day from the past data; a third step of integrating the clear weather power generation curve to obtain an annual transition curve showing the annual transition of the total power generation amount obtained; a fourth step of scaling the clear weather power generation curve to match the annual progression curve to obtain a clear weather power generation model; a fifth step of obtaining the clear-sky power generation model for all days of the year; A sixth step of obtaining weather coefficients corresponding to each weather; A method for predicting power generation, characterized by executing the steps of: obtaining a weather forecast for the target day; and a seventh step of obtaining a predicted power generation output from the power conditioner from the clear weather power generation model and the weather coefficients for the target day.

2. 2. The method for predicting power generation according to claim 1, wherein the clear weather power generation curve is expressed by a function that is symmetrical with respect to a time axis at the center of power generation time.

3. 3. The method for predicting power generation according to claim 1, wherein in the second step, the selected day has a total sunny time of 40% or more of the total power generation time.

4. 4. The method for predicting power generation according to claim 1, wherein the weather coefficients are defined for each hour or month.

5. In the second step, The clear weather power generation curve is calculated based on the power generation value at the time when the power generation value becomes equal to or less than the rated output of the power conditioner; A method for predicting power generation according to any one of claims 1 to 4, characterized in that in the seventh step, the minimum value of the predicted power generation calculated for each hour from the clear weather power generation model and the weather coefficient and the rated output is adopted as the predicted power generation for that hour.

6. 5. A method for predicting power generation according to claim 1, further comprising the step of correcting the predicted power generation after the target prediction date by comparing the power generation output from the power conditioner actually measured on the target prediction date with the predicted power generation.

7. A solar power generation system including a solar cell, a power conditioner, and a power generation prediction device, The photovoltaic power generation system, wherein the power generation prediction device executes the prediction method according to any one of claims 1 to 6.

8. A power generation prediction device for predicting power generation in a photovoltaic power generation system including a solar cell and a power conditioner, a processor and a storage device; 7. A power generation prediction device, wherein the arithmetic processing unit executes the prediction method according to claim 1.