KR102491085B1 - Apparatus and Method for Predicting Generated Power of Wind Power Generator based on Atmospheric Model Forecast - Google Patents

Abstract

The present invention is a meteorological information collection unit that collects meteorological information including meteorological forecasts, receives a meteorological forecast of low resolution temporally and spatially, and numerically analyzes the terrain altitude and surrounding environmental factors according to the installation location where the wind turbine is installed to reflect turbulence spread Atmospheric model application unit that applies an atmospheric model to generate high resolution by applying an atmospheric model that generates a plurality of time variables that continuously repeat and increase or decrease at a predetermined period, respectively, and generate time variable values according to the generated plurality of time variables into a plurality of high-resolution weather conditions A time variable application unit that generates input data by matching weather information including forecasts, and input data that is pre-learned based on learning data in which previously acquired input data and actually generated power generation are matched to generate input data, and is currently applied according to the learned method Provided is a wind power generation forecasting device and method capable of predicting wind power generation with high accuracy, including a power generation predictor for estimating and outputting the power generation amount from.

Description

Apparatus and Method for Predicting Generated Power of Wind Power Generator based on Atmospheric Model Forecast}

The present invention relates to an apparatus and method for predicting wind power generation, and relates to an apparatus and method for predicting wind power generation based on atmospheric model forecast using an artificial neural network.

Demand for eco-friendly energy sources is rapidly increasing in accordance with the global trend of low-carbonization, and as a result, interest in eco-friendly renewable energy such as wind power and solar power generation is also greatly increasing. Among these new and renewable energies, wind power generation has advantages in that it can be installed on a small scale and has a low power generation cost. However, wind power generation is a power generation method that uses wind to produce power, so the amount of power generation varies depending on the strength of the wind, that is, the wind speed.

Such characteristics of wind power generation make it difficult to predict or control the amount of power generation compared to power generation methods in which the amount of power generation is easy to control, such as thermal power or nuclear power, making it difficult to efficiently manage power. Accordingly, various studies have been conducted to predict wind power generation based on weather forecasts. However, the wind direction and speed provided by the weather forecast are forecast values for a long time period and a wide area, and their spatio-temporal resolution is low, making it difficult to represent the wind speed at the place where the wind turbine is installed. there is.

Korean Registered Patent No. 10-1844044 (registered on March 26, 2018)

An object of the present invention is to provide a wind power generation forecasting device and method capable of accurately predicting the power generation of a wind power generator by applying a time variable for imparting temporal continuity to weather forecast data input to an artificial neural network.

Another object of the present invention is to provide a wind power generation forecasting device and method capable of more accurately predicting the power generation of a wind power generator by obtaining wind speed information of high time and space resolution using an atmospheric model from various weather forecast data provided at low time and space resolution. there is.

An apparatus for predicting wind power generation according to an embodiment of the present invention for achieving the above object includes a weather information collection unit for collecting weather information including a weather forecast; Atmospheric model applicator for receiving the weather forecast of temporally and spatially low resolution and applying an atmospheric model that numerically analyzes the topographical altitude and surrounding environmental factors according to the installation location where the wind power generator is installed to the turbulence diffusion to achieve higher resolution; Time to generate input data by generating a plurality of time variables that continuously increase or decrease at predetermined intervals, and matching time variable values according to the generated plurality of time variables with weather information including a plurality of high-resolution weather forecasts. variable application unit; and a power generation prediction unit that is pre-learned based on previously obtained input data and actual generated power matching learning data, and estimates and outputs power generation from currently applied input data according to the learned method.

The time variable application unit generates two date variables having different offsets that continuously increase or decrease in a one-year cycle and two time variables that continuously increase or decrease in a one-day cycle and have different offsets, and calculates the amount of power generation. A time variable value for a time to be predicted may be matched with the weather information.

The atmospheric model applying unit calculates a bulk Richardson number (Bi _b ), which is an indicator of the dynamic stability of the atmosphere, and an atmospheric stability calculation unit for calculating atmospheric stability (z _r /L) using the bulk Richardson number (Bi _b ). ; A zero-plane displacement calculation unit that calculates a stability variable (β) and a zero-plane displacement height (d _t ), which is a ground height newly defined in the atmospheric model, based on a canopy height (h), which is a height of an obstacle on the ground obtained by measurement; a roughness length calculation unit that calculates a roughness length (z ₀ ) representing a height that becomes 0 when it is assumed that the wind follows a logarithmic profile; and friction velocity (u _* ) is calculated using the calculated atmospheric stability (z _r /L) and roughness length (z ₀ ), and the wind speed (C _* ) above the canopy height (h) and the canopy height (h) A vertical wind speed determiner may be included to obtain a vertical wind speed profile by calculating the wind speed (C _* ) below.

The atmospheric model application unit obtains a standard deviation for a subgrid-scale orography shape of a predetermined size based on the location where the wind power generator is installed, and is located in the horizontal and vertical directions around a specific sub-grid. The Laplacian value of terrain elevation is obtained based on the height relationship with four surrounding sub-grids, and different winds are obtained according to the Laplacian value of terrain altitude according to the relationship between the standard deviation of the mountain shape of the sub-grid scale and the surrounding topography. It may further include a topography-based wind speed correction unit that additionally reflects the dissipation level of the vertical wind speed profile obtained by the vertical wind speed determining unit.

The weather information collection unit based on the location where the wind power generator is installed, a forecast guarantee collection unit for collecting a forecast report predicting a weather condition at a predetermined time interval thereafter; A nacelle data collection unit for collecting real-time meteorological information measured using a measurement device installed in the nacelle of the wind power generator may be included.

In order to achieve the above object, a wind power generation prediction method according to another embodiment of the present invention includes collecting weather information including a weather forecast; Applying an atmospheric model that numerically analyzes the terrain altitude and surrounding environmental factors according to the installation location where the wind power generator is installed by receiving the weather forecast of low resolution in time and space to the turbulent spread to obtain high resolution; Generating a plurality of time variables that continuously increase or decrease at predetermined intervals, and generating input data by matching time variable values according to the generated plurality of time variables with weather information including a plurality of high-resolution weather forecasts. ; and pre-learning based on learning data in which previously obtained input data and actual generated power are matched, and estimating and outputting the power generation from currently applied input data according to the learned method.

Therefore, the apparatus and method for predicting wind power generation according to an embodiment of the present invention obtain wind speed information with high spatial-temporal resolution using an atmospheric model from various weather forecast data provided with low spatial-temporal resolution, and obtain temporal and temporal information on the obtained high-resolution wind speed information. Wind power generation can be predicted with high accuracy by applying a time variable for imparting continuity and inputting it to an artificial neural network.

1 shows a schematic configuration of a wind power generation prediction device according to an embodiment of the present invention.
FIG. 2 shows an example of a detailed configuration of the weather forecast collection unit of FIG. 1 .
3 is a diagram for explaining the planetary boundary layer according to altitude.
FIG. 4 is a view for explaining a vertical coordinate system used in the atmospheric model application unit of FIG. 1 .
5 shows an example of a detailed configuration of the atmospheric model application unit of FIG. 1 .
6 shows an example of a vertical wind profile according to the atmospheric model.
7 is a diagram for explaining a terrain effect affecting wind speed.
8 shows an example of a high-resolution forecast obtained by the atmospheric model application unit.
FIG. 9 shows an example of a time variable generated by the time variable generator of FIG. 1 .
10 shows an example of input data input to the generation amount prediction unit.
11 shows a wind power generation prediction method according to an embodiment of the present invention.

In order to fully understand the present invention and its operational advantages and objectives achieved by the practice of the present invention, reference should be made to the accompanying drawings illustrating preferred embodiments of the present invention and the contents described in the accompanying drawings.

Hereinafter, the present invention will be described in detail by describing preferred embodiments of the present invention with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the described embodiments. And, in order to clearly describe the present invention, parts irrelevant to the description are omitted, and the same reference numerals in the drawings indicate the same members.

Throughout the specification, when a part "includes" a certain component, it means that it may further include other components, not excluding other components unless otherwise stated. In addition, terms such as "... unit", "... unit", "module", and "block" described in the specification mean a unit that processes at least one function or operation, which is hardware, software, or hardware. and software.

1 shows a schematic configuration of an apparatus for predicting wind power generation amount according to an embodiment of the present invention, and FIG. 2 shows an example of a detailed configuration of a weather forecast collection unit of FIG. 1 .

Referring to FIG. 1 , the apparatus for predicting wind power generation according to the present embodiment may include an information collection unit 100, an atmospheric model application unit 200, a time variable application unit 300, and a power generation estimation unit 400. .

The information collection unit 100 collects various meteorological information that can be collected about the location where the wind power generator is installed. In this embodiment, the information collection unit 100 not only collects and obtains a weather forecast according to the location information of the wind power generator, but additionally provides information on the weather conditions currently measured in the wind power generator and the weather conditions previously measured in the vicinity. information can be gathered together.

The information collection unit 100 basically collects weather forecasts as information for predicting wind power generation. However, it is obvious that more weather information can improve the accuracy of forecasting. Accordingly, the information collection unit 100 may further collect currently measured weather information and past actually measured weather information. That is, it is possible to improve the accuracy of forecasting wind power generation by additionally collecting not only the weather forecast obtained through future forecasting but also current and past actually measured weather information.

In addition, the information collection unit 100 collects the amount of power currently generated by the wind power generator together with weather information.

Referring to FIG. 2 , the information collection unit 100 may include a forecast guarantee collection unit 110, a nacelle data collection unit 120, and an AWS data collection unit 130.

The forecast collection unit 110 collects weather forecasts provided in a predetermined manner. The forecast guarantee collection unit 110 collects and obtains forecast guarantees (hereinafter referred to as GFS forecast guarantees) provided by the Global Forecast System (GFS) of the National Centers for Environmental Prediction (NCEP) as an example. can

The nacelle data collection unit 120 collects meteorological information measured in the nacelle of the wind power generator. In a wind power generator, a nacelle in which a rotor rotating by the wind is coupled and a generator that generates power by being connected to the rotating shaft of the rotor is disposed therein is generally equipped with an anemometer and a wind vane as measuring devices, which are actually wind power. It is used as a means to most accurately measure the strength and direction of the wind blowing into the generator, that is, the wind speed and direction in real time. Therefore, the nacelle data collection unit 120 collects meteorological information measured in real time in the nacelle of the wind power generator to predict the amount of power generation. In addition, the nacelle data collection unit 120 detects and collects the amount of power actually generated by the wind power generator.

That is, the nacelle data collection unit 120 may collect information about the wind speed actually blowing to the wind power generator and the amount of power generated by the wind power generator accordingly.

Meanwhile, the AWS data collection unit 130 collects and obtains weather information measured by an Automatic Weather Station (hereinafter referred to as AWS) located at a short distance from the wind power generator. AWS is a device that is installed in a pre-designated location and constantly observes weather changes at that location. That is, the AWS can obtain meteorological information actually observed around the wind power generator. However, since the weather information collected by AWS is not provided in real time, but is typically provided as information acquired the previous day, the AWS data collection unit 130 may collect ambient weather information obtained the previous day. Here, the AWS data collection unit 130 is a component added to improve the accuracy of power generation prediction and may be omitted in some cases.

Meanwhile, the atmospheric model application unit 200 temporally and spatially high-resolution weather forecasts among the weather information collected by the information collection unit 100 . Typically, weather forecasts are provided over a large area at long time intervals. For example, the GFS forecast is provided at 00, 06, 12, and 18 UTC (09, 15, 21, and 03 KST), and weather forecasts for 16 days up to 384 hours after a 3-hour interval are provided. In addition, the spatial resolution of the weather forecast provided by the GFS forecaster is approximately 25 km. That is, the GFS forecaster provides weather information on a scale of 25 km at a time interval of 3 hours, which can be seen as having low resolution temporally and spatially.

Therefore, the weather forecast has a significant error with the actual weather condition at the time of predicting the location and power generation of the wind turbine, and thus, the prediction of the wind power generation also has a very large error. In particular, since most of the wind power generators are installed densely in mountainous or marine areas, errors in forecasting appear even larger. Accordingly, the atmospheric model application unit 200 constructs an atmospheric model to obtain a weather forecast suitable for the location where the wind power generator is installed and the time to predict the amount of power generation in order to reduce the prediction error due to the low resolution of the weather forecast. High-resolution weather forecast.

At this time, the atmospheric model application unit 200 may increase the low-resolution weather forecast to high resolution by applying an atmospheric model reflecting the atmospheric conditions according to the height at which the wind power generator is installed and the surrounding topography. For example, the atmospheric model application unit 200 may high-resolution meteorological information having a time interval of 3 hours and a spatial interval of 25 km to have an interval of 10 minutes and a spatial interval of 500 m.

3 is a diagram for explaining a planetary boundary layer according to altitude, FIG. 4 is a diagram for explaining a vertical coordinate system used in the atmospheric model application unit of FIG. 1, and FIG. 5 is a detailed configuration of the atmospheric model application unit of FIG. 1 indicate an example. And Figure 6 shows an example of a vertical wind profile according to the atmospheric model.

Referring to FIG. 3, the lowermost layer of the atmosphere at a height of about 1 to 2Km from the ground surface, and the atmospheric layer in which air currents interact due to friction with the ground is called the Planetary Boundary Layer (PBL), and the lower 10% in PBL The height of the degree (about 100 ~ 200m) is called the surface boundary layer (SL) (or referred to as the surface layer) that directly interacts with the ground. SL is a very important area in atmospheric modeling where ground influences such as friction, atmospheric heating and moisture supply are directly applied, and atmospheric models for PBL or SL have been previously proposed.

And, as shown in FIG. 3 again, the SL is divided into an upper inertial sublayer (hereinafter referred to as ISL) and a lower roughness sublayer (hereinafter referred to as RSL). In the case of the ISL located in the upper part of the two, the MOST theory (Monin-Obukhov Similarity Theory), which describes the vertical distribution of wind and temperature in the stationary turbulent state in the SL, is applied to The atmospheric state considering the influence of can be easily modeled. On the other hand, since the RSL located at the bottom is directly affected by obstacles on the ground, the existing MOST theory alone cannot accurately model the atmospheric state, that is, the wind speed. However, since the actual wind turbine is located at the RSL, there is a limitation that the wind speed applied to the wind turbine cannot be simulated in the conventional method.

Accordingly, the atmospheric model application unit 200 of this embodiment parameterizes the atmospheric state in the RSL, that is, the turbulence diffusion, with a new modeling technique, so that the wind speed applied to the actual wind power generator can be accurately simulated at high spatial and temporal resolution at high resolution. . The atmospheric model application unit 200 of the present embodiment may represent a more accurate wind speed than the WRF (Weather Research and Forecasting) model, which is a mesoscale model previously developed by the National Center for Atmospheric Research (NCAR) to numerically analyze the atmosphere. It is modified so that it can be used as an atmospheric model.

Referring to FIG. 4 , the atmospheric model application unit 200 of the present embodiment uses a vertical coordinate system that is re-corrected as in (a) from the vertical coordinate system proposed as in (b) to numerically analyze the atmospheric state of the RSL.

= 0로 표현되고 d₀ 는 기존에 WRF 모델 내부의 MM5 SL 체계에서 정의된 영면 변위 높이(zero-plane displacement height)를 나타내며, z₀는 장애물에 의한 거칠기 길이(roughness length), h는 장애물 높이를 나타내는 캐노피 높이(canopy height), z_r은 기존 WRF 모델의 좌표계에서 최하 계층(여기서는 RSL)의 길이로서, 영면 변위 높이(d₀)로부터의 거칠기 길이를 나타낸다.In Figure 4, the ground surface is

= 0, d ₀ represents the zero-plane displacement height previously defined in the MM5 SL system inside the WRF model, z ₀ is the roughness length due to the obstacle, and h is the height of the obstacle The canopy height representing , z _r is the length of the lowest layer (here, RSL) in the coordinate system of the existing WRF model, and represents the roughness length from the zero plane displacement height (d ₀ ).

)에서 영면 변위 높이(

= d₀)를 수직 좌표계(z)의 원점으로 설정하여 대기 모형을 적용한다. 따라서 영면 변위 높이(

= d₀)가 수직 좌표계(z)의 원점으로 적용됨에 따라, 본 실시예에서 수직 좌표계(z)는 z =

- d₀ 이다. 그리고 수정된 수직 좌표계(z)를 기반으로 하만 및 피니간(Harman and Finnigan(2007): 이하 HF)에 의해 재정의된 영면 변위 높이(d_t)는 캐노피 높이(h)에서 기존 영면 변위 높이(d₀)를 차감한 값(d_t = h - d₀)으로 계산될 수 있다. 이하에서는 도 4에서 (a)의 수직 좌표계를 이용하는 것으로 가정한다.Comparing (a) and (b), the atmospheric model application unit 200 according to the present embodiment, as shown in (a), the existing vertical coordinate system (

) at the zero plane displacement height (

= d ₀ ) is set as the origin of the vertical coordinate system (z) to apply the atmospheric model. Therefore, the zero plane displacement height (

= d ₀ ) is applied as the origin of the vertical coordinate system (z), in this embodiment, the vertical coordinate system (z) is z =

-d is ₀ . And, based on the modified vertical coordinate system (z), the zero plane displacement height (d _t ) redefined by Harman and Finnigan (Harman and Finnigan (2007): hereinafter HF) is the existing zero plane displacement height (d ₀ ) subtracted (d _t = h - d ₀ ). Hereinafter, it is assumed that the vertical coordinate system of (a) in FIG. 4 is used.

Referring to FIG. 5, the atmospheric model application unit 200 includes an atmospheric stability calculation unit 210, a zero plane displacement calculation unit 220, a tree height measurement unit 230, a roughness length calculation unit 240, a vertical wind speed determination unit ( 250) and a terrain-based wind speed correction unit 260.

The atmospheric stability calculator 210 calculates atmospheric stability (z _r /L) as one of the variables necessary for calculating the vertical wind speed. The atmospheric stability calculator 210 calculates a Bulk Richardson Number (Bi _b ) prior to calculating the atmospheric stability (z _r /L). The bulk Richardson number (Bi _b ) is an indicator of the dynamic stability of the atmosphere and is used to determine whether turbulence occurs. The bulk Richardson number can be calculated according to Equation 1.

where g is the gravitational acceleration, and θ _a is the potential temperature at the height (z _r ) of the lowest model layer (here, RSL) in the coordinate system of the existing WRF model (MM5 SL) of the numerical model , and θ _va and θ _vg represent the virtual potential temperature at SL and ground surface, respectively. And u(z _r ) represents the wind speed at the height of the lowest model layer (z _r ), and z is a vertical coordinate value with the zero plane displacement height (d ₀ ) defined in the existing WRF model (MM5 SL) as the origin.

When the bulk Richardson number is calculated, the atmospheric stability calculator 210 calculates the atmospheric stability (z _r /L) by applying the calculated bulk Richardson number. Atmospheric stability (z _r /L) is the ratio of the model bottom height (z _r ) to the obukov length (L). The obukov length (L) itself represents stability, but here reflects the model bottom height (z _r ). It scales dimensionlessly and is calculated according to Equation 2.

,

는 각각 운동량과 열에 관한 적분된 RSL 함수를 나타낸다. 그리고 ρ는 공기 밀도이고, c_p 는 공기에 대한 정압비열(Specific heat for air)이며, c_s 는 열 전달 효율 계수(Effective heat transfer coefficient)를 나타낸다. z₀, z_l 은 각각 거칠기 길이와 비스커스 아층 깊이(Viscous sublayer depth)이고, k는 본 카르만 상수(von Kㅱrmㅱn constant)(여기서는 0.4)이다.where u _* ^n-1 represents the friction velocity obtained in the previous step (n-1), ψ _m and ψ _h represent the integrated similarity function for momentum and heat, respectively,

,

denotes the integrated RSL function for momentum and heat, respectively. And ρ is the air density, c _p is the specific heat for air, and c _s represents the effective heat transfer coefficient. z ₀ and z _l are the roughness length and the viscous sublayer depth, respectively, and k is the von Karman constant (here 0.4).

,

)를 추가한 구성으로, RSL 함수(

,

)는 HF에 의해 관측을 기반으로 경험적으로 획득된 알려진 함수이다.Equation 2 is an RSL function (to apply the existing MOST theory to RSL)

,

) with the addition of the RSL function (

,

) is a known function obtained empirically based on observations by HF.

The zero plane displacement calculation unit 220 calculates the height of the ground displacement (d _t ) that is newly defined in consideration of obstacle elements on the ground.

At this time, the zero plane displacement calculation unit 220 may obtain the tree height measured by the tree height measurement unit 230 using Lidar or the like as the canopy height h. In general, wind power generators must be able to blow constantly, so they are installed on highlands or on the sea, so trees are most of the surrounding obstacles rather than buildings. Therefore, it is assumed here that the tree height measured by the tree height measuring unit 230 is used as the canopy height h. In some cases, the height of the surrounding trees may be already measured and stored in the zero plane displacement calculation unit 220 in advance. In this case, the tree height measurement unit 230 may be omitted.

The zero plane displacement calculator 220 may calculate the zero plane displacement height d _t according to Equation 3.

Here, l _m is the mixing length for momentum, β is the stability parameter, and L _c is the canopy penetration depth.

In Equation 3, the canopy penetration depth (L _c ) and the stability parameter (β) are calculated according to Equations 4 and 5, respectively.

where c _d is the drag coefficient at the leaf scale, a is the leaf area density, and LAI is the leaf area index.

Here, Φ _m is a similar function related to momentum, and β _N represents the stability value in the natural state (=0.374).

That is, the zero-plane displacement calculation unit 220 calculates and obtains the zero-plane displacement height (d _t ), which is the ground height newly defined in the atmospheric model, in consideration of the tree height, which is an obstacle element on the ground surface.

,

)에도 영향을 미치는 변수이다. 따라서 대기 안정도 계산부(210)와 영면 변위 계산부(220)는 안정도 변수(β)와 RSL 함수(

,

)의 변화가 기지정된 기준값 이하가 될 때까지 상호 반복하여 대기 안정도(z_r/L)와 안정도 변수(β)를 교대로 계산한 후, 영면 변위 높이(d_t)를 계산할 수 있다.However, the stability variable (β) is the RSL function (

,

) is also a variable that affects Therefore, the atmospheric stability calculator 210 and the zero plane displacement calculator 220 use the stability variable β and the RSL function (

,

) is mutually repeated until the change in ) is less than or equal to a predetermined reference value, and the atmospheric stability (z _r /L) and the stability variable (β) are alternately calculated, and then the zero surface displacement height (d _t ) can be calculated.

The roughness length calculation unit 240 calculates a roughness length (z ₀ ) representing a height at which the wind becomes zero, assuming that the wind follows a logarithmic profile as shown in FIG. 6 . The roughness length z ₀ shown in FIG. 4 may be calculated according to Equation 6.

은 운동량에 관한 RSL 상사 함수를 나타낸다.here

denotes the RSL similar function for momentum.

The vertical wind speed determiner 250 calculates a friction velocity (u _* ) according to the vertical coordinate value (z) according to Equation 7 to determine the vertical wind profile.

Wind speed (u) or temperature (T) may be calculated by substituting a variable (C) when calculating the vertical profile. However, the vertical profile for variable (C) moves up and down based on the canopy height (h) according to the relative distance between the canopy height (h) and the height of the zero plane displacement (d _t = h - d ₀ ) redefined by HF. Separately, the canopy height (h) or more is calculated by Equation 8, and the canopy height (h) or less can be calculated by Equation 9.

는 변수(c)의 적분 상사 함수(integrated similarity function)이며,

는 변수(C)의 RSL 함수이다.where C _* is the scale value of variable (C), C ₀ is the value of variable (c) at the roughness length (z ₀ ),

is the integrated similarity function of the variable (c),

is the RSL function of the variable (C).

where C _h is the value of variable (c) at canopy height (h), and f is a parameter related to the depth scale of the scalar profile.

In this embodiment, when the variable C is the wind speed, as shown in FIG. 6, the vertical wind speed profile of a pattern divided differently from top to bottom based on the canopy height (h) can be obtained. .

Compared to the case where the conventional MOST theory is applied, the wind speed according to the terrain altitude can be more accurately expressed in the RSL where the wind turbine is located.

When the vertical wind speed profile is obtained according to Equations 8 and 9, the terrain-based wind speed correction unit 260 additionally corrects the wind speed by reflecting the terrain characteristics of the location where the wind turbine is installed. The wind speed according to the vertical wind speed profile shows a trend depending on the topography. Generally, the wind speed tends to be under-simulated at the top of a mountain, whereas it tends to be over-simulated in a plain or valley.

Therefore, in order to more accurately calculate the wind speed of the wind applied to the wind power generator, the terrain-based wind speed corrector 260 additionally compensates for the wind speed obtained from the vertical wind speed profile according to the terrain characteristics.

Here, the wind speed correction obtains the standard deviation (σ _sso ) for a subgrid-scale orography shape of a predetermined size, and four surrounding sub-grids located in the horizontal and vertical directions centered on a specific sub-grid A height relationship with is calculated according to Equation 10 to obtain a Laplacian value (Δ ² h _i,j ) of terrain elevation.

In addition, the wind speed due to the friction of the ground, that is, the wind speed according to the vertical wind speed profile, is compensated by reflecting the dissipation level as shown in Equation 11.

At this time, the dissipation factor (c _t ) representing the dissipation level is based on the standard deviation (σ _sso ) of the mountain topography of the sub-grid scale and the Laplacian value (Δ ² h) of the terrain elevation according to the relationship with the surrounding topography. Equation Obtained according to 12.

In Equation 12, the dissipation factor (c _t ) has a value of 1 at the top at the plain location, so that the wind speed is greatly reduced, thereby reducing the excessively simulated wind speed, whereas at the top of the mountain, it becomes 0 and the wind speed is not reduced. prevent mocking

In addition, in mountainous terrain, the top of a mountain or the slope where sunlight reaches is heated by the sun's heat, while the shadow of the mountain forms a shaded area in the plain or valley to maintain a relatively low temperature, resulting in wind changes according to the terrain additionally occurs.

7 is a diagram for explaining a terrain effect affecting wind speed.

As shown in FIG. 7 , the terrain-based wind speed correction unit 260 sets the solar radiation to be 0 in the shadow area based on the position of the sun over time and the mountainous topography, and adjusts the sun radiation according to the slope of the terrain in the inclined area where the sun shines. The amount of solar radiation is reflected in each of the warming levels (θ _va , θ _vg ) in Equation 1.

The topography-based wind speed corrector 260 may weight Equation 13 so that the amount of solar radiation according to the angle θ of the solar altitude relative to the slope is reflected in each of the warming positions θ _va and θ _vg in the slope region.

Here, φ is the earth's latitude, δ is the declination (declination), β is the inclination of the earth's surface, γ is the azimuth angle, ω is the angle formed by the meridian and the celestial body, and θ is the angle of incidence.

As described above, the atmospheric model application unit 200 may extract a high-resolution weather forecast from a weather forecast having a temporal and spatial low resolution using an atmospheric model that is a numerical analysis model.

8 shows an example of a high-resolution forecast obtained by the atmospheric model application unit.

In FIG. 8, (a) and (b) show the GFS collected in January and February, respectively, and the atmospheric model application unit 200 applies the atmospheric model to generate the GFS at the location of the wind power plant It shows the result of comparing the high-resolution result (WRF-GFS) and the wind speed actually measured at the forecasted time (observed). As shown in FIG. 8, the high-resolution wind speed (WRF-GFS) by the atmospheric model application unit 200 according to this embodiment appears in a pattern very close to the actually measured wind speed compared to the forecast (GFS). Able to know.

Meanwhile, the time variable application unit 300 uses a predetermined method to reflect the weather information collected by the information collection unit 100 and the change in successive times in the high-resolution weather forecast by the atmospheric model application unit 200. Create and apply to generate input data. A detailed description of the time variable application unit 300 will be described later.

The power generation prediction unit 400 is implemented as a pre-trained artificial neural network, receives input data that is weather information to which a time variable is applied, and estimates and outputs power generation according to the learned method.

The weather database 500 stores learning data for learning the generation amount predictor 400 . The weather database 500 accumulates and stores input data previously obtained by the information collection unit 100 and applying the time variable to the time variable application unit 300 and actual power generation at the time according to the time variable as learning data.

Accordingly, the power generation prediction unit 400 receives the input data accumulated and stored in the weather database 500 to estimate the power generation amount, and the difference between the estimated power generation amount and the actual power generation amount stored in the weather database 500 corresponding to the input data is By being propagated back as a loss, it can be learned to accurately estimate the amount of power generation corresponding to the input data.

FIG. 9 shows an example of a time variable generated by the time variable generator of FIG. 1 , and FIG. 10 shows an example of input data input to the generation amount predictor.

As shown in FIG. 9, the time variable application unit 300 uses two date variables (var1 and var2) representing the date of month and day and two time variables (var3 and var4) representing the time of hours and minutes. can create

Various weather information has a time to be collected, and in the case of a weather forecast, a forecasted time also exists. As described above, in the case of the S forecast report collected by the forecast report collection unit 110, 16-day weather forecasts are provided at 00, 06, 12, and 18 UTC for up to 384 hours after a 3-hour interval. However, since the weather forecast increases in error as the future is predicted, it is meaningless to increase the resolution of the entire forecast provided by the atmospheric model application unit 200. Accordingly, the atmospheric model application unit 200 may increase resolution only for forecasts up to a predetermined period (24 hours thereafter).

Meanwhile, unlike a weather forecast, measurement values collected by the nacelle data collection unit 120 and the AWS data collection unit 130 may periodically collect current or previously measured wind speed data at designated time intervals. Therefore, the weather information collected by the weather forecast collector 110, the weather information collected by the nacelle data collector 120, and the weather information collected by the AWS data collector 130 may represent weather information for different times. there is.

However, in this embodiment, the time variable application unit 300 creates the time variable as shown in FIG. 9 and matches weather information collected at corresponding times to the time to predict the amount of power generation to generate input data as shown in FIG. 10. can That is, input data may be generated by matching currently collected weather information to a time variable to predict power generation regardless of the time indicated by the weather information included in the collected weather information. At this time, in the case of a low-resolution weather forecast, input data is generated by matching the high-resolution weather forecast to a time variable by the atmospheric model application unit 200 .

However, when the time information used by actual people is used as a time variable, learning of the power generation predictor 400 is not easily performed, and there is a high possibility that an error will occur in the power generation prediction even after learning. It was confirmed that this is because the time information according to the generally used time display system is discontinuously expressed. That is, January after December in the year unit, and 1 day after the 30th or 31st day in the unit of month, and 1 hour after 12:00 and 1 minute after 60 minutes are repeated discontinuously in the unit of hours and minutes, respectively. Errors may occur because the information on is not easily processed.

Accordingly, in this embodiment, as shown in FIG. 9 , the time variable application unit 300 generates a date variable that repeatedly increases or decreases with a one-year cycle and a time variable that repeatedly increases or decreases with a daily cycle, and matches them to the input data, The generation amount estimation unit 400 implemented as an artificial neural network can easily accommodate time information. However, if a date variable and a time variable are implemented as one variable each, the continuity of date and time can be provided. do. To prevent this problem, the time variable application unit 300 generates two date variables (var1, var2) and two time variables (var3, var4) that increase or decrease at the same cycle, and two date variables (var1, var2) has different offset values and two time variables (var3, var4) also have different offset values, so that the date and time can be easily distinguished while ensuring the continuity of time.

Therefore, as shown in FIG. 10, the input data input to the power generation prediction unit 400 is matched with weather information and time variables ((var1, var2), (var3, var4) representing the date and time for predicting the power generation. At this time, the weather information may include at least one of nacelle data and AWS data together with high-resolution forecast data.

11 shows a wind power generation prediction method according to an embodiment of the present invention.

Referring to FIGS. 1 to 10, the wind power generation prediction method according to the present embodiment is described. First, various weather information and power generation information are collected (S10). At this time, the weather information collects a weather forecast and other collectable weather information based on the location where the wind power generator is installed. Weather forecasts can then be collected up to a predetermined period of time (24 hours as an example in this case), and other weather information includes the wind speed currently measured by the measuring device installed in the nacelle of the wind power generator and the wind speed previously measured by the adjacent ASW. etc. may be included.

Then, among the collected meteorological information, an atmospheric model that numerically reflects the topographical altitude and surrounding environmental factors according to the installation location of the wind power generator to the turbulence spread is applied to the weather forecast with low spatial and temporal resolution to achieve high resolution (S20).

In the step of upgrading the resolution, first, the bulk Richardson number (Bi _b ), which is an indicator of the dynamic stability of the atmosphere, is calculated according to Equation 1, and the atmospheric stability (z _r /L) is calculated using the calculated bulk Richardson number (Bi _b ). is calculated according to Equation 2 (S21).

,

) 또는 안정도 변수(β)의 변화가 기지정된 기준값 이하인지 판별한다(S23). 만일 대기 안정도(z_r/L)의 RSL 함수(

,

) 또는 안정도 변수(β)의 변화가 기준값을 초과하는 것으로 판별되면, 계산된 안정도 변수(β)를 기반으로 RSL 함수(

,

)를 업데이트하여 대기 안정도(z_r/L)를 다시 계산하고(S21), 영면 변위 높이(d_t)를 업데이트한다(S22).Based on the stability variable (β) and the canopy height (h), which is the height of trees measured using LIDAR, the height of the ground displacement (d _t ), which is newly defined in the atmospheric model, is calculated according to Equation 3 ( S22). And the RSL function of atmospheric stability (z _r /L) (

,

) or whether the change in the stability variable (β) is less than or equal to a predetermined reference value (S23). If the RSL function of atmospheric stability (z _r /L) (

,

) or if it is determined that the change in the stability variable (β) exceeds the reference value, the RSL function (

,

) is updated to recalculate the atmospheric stability (z _r /L) (S21), and the zero plane displacement height (d _t ) is updated (S22).

,

) 또는 안정도 변수(β)의 변화가 기준값 이하이면, 바람이 로그 프로파일을 따른다고 가정할때, 0이 되는 높이를 나타내는 거칠기 길이(z₀)를 수학식 6에 따라 계산한다(S24). 그리고 계산된 대기 안정도(z_r/L)와 거칠기 길이(z₀)를 이용하여 마찰 속도(u_*)를 수학식 7에 따라 계산한다(S25).However, the RSL function of atmospheric stability (z _r /L) (

,

) or if the change in the stability variable (β) is less than or equal to the reference value, assuming that the wind follows a logarithmic profile, the roughness length (z ₀ ) representing the height of 0 is calculated according to Equation 6 (S24). Then, the friction speed (u _* ) is calculated according to Equation 7 using the calculated atmospheric stability (z _r /L) and the roughness length (z ₀ ) (S25).

Then, based on the canopy height (h), the wind speed (C _{* ) above the canopy height (h) and the wind speed (C * )} below the canopy height (h) are calculated according to Equations 8 and 9, respectively, and the vertical wind speed A profile is acquired (S26). In addition, the wind speed according to the terrain elevation of the calculated vertical profile is corrected according to the effect of the terrain where the wind turbine is installed (S27). Here, the correction according to the terrain effect obtains the standard deviation (σ _sso ) of the mountain topography on a sub-grid scale and the Laplacian value (Δ ² h _i,j ) of the altitude of the surrounding terrain, and calculates the dissipation level of the wind speed according to the terrain by the equation 11 can be reflected and compensated.

In some cases, when calculating the bulk Richardson number (Bi _b ), the weight of solar radiation according to the slope of the terrain where the wind turbine is installed and whether it is sunny or shaded is reflected in advance to compensate for wind speed.

On the other hand, when weather information is obtained and the low-resolution weather forecast is applied to the atmospheric model to achieve high resolution, the input data is obtained by matching the high-resolution weather forecast and the remaining weather information with time variables generated in response to the time to predict the amount of power generation. Obtain (S30). At this time, the time variable can be created as two date variables (var1, var2) with different offsets that increase or decrease in a cycle of one year and two time variables (var3, var4) with different offsets that increase or decrease in a cycle of one day. .

Then, the obtained input data is input to the pre-learned artificial neural network to predict the amount of power generation of the wind power generator at a time specified as a time variable. Here, the artificial neural network is pre-learned based on input data in which time variables are matched to previously accumulated weather information and learning data including the amount of power actually generated and measured by the wind power generator in response thereto. The artificial neural network may be learned by backpropagating as a loss an error between the amount of power generation estimated by receiving the input data of the learning data and the amount of power generation actually generated and measured by the wind power generator.

The method according to the present invention may be implemented as a computer program stored in a medium for execution on a computer. Here, computer readable media may be any available media that can be accessed by a computer, and may also include all computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, including read-only memory (ROM) dedicated memory), random access memory (RAM), compact disk (CD)-ROM, digital video disk (DVD)-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

Although the present invention has been described with reference to the embodiments shown in the drawings, this is only exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom.

Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

100: information collection unit 200: atmospheric model application unit
210: Atmospheric stability calculation unit 220: Zero plane displacement calculation unit
230: tree height measurement unit 240: roughness length calculation unit
250: vertical wind speed determination unit 260: terrain-based wind speed correction unit
300: time variable application unit 400: power generation estimation unit
500: weather database

Claims (20)

a weather information collection unit that collects weather information including weather forecast;
Atmospheric model applicator for receiving the weather forecast of temporally and spatially low resolution and applying an atmospheric model that numerically analyzes the topographical altitude and surrounding environmental factors according to the installation location where the wind power generator is installed to the turbulence diffusion to achieve higher resolution;
Time to generate input data by generating a plurality of time variables that continuously increase or decrease at predetermined intervals, and matching time variable values according to the generated plurality of time variables with weather information including a plurality of high-resolution weather forecasts. variable application unit; and
Wind power generation forecasting device including a power generation estimation unit that is pre-learned based on learning data in which previously obtained input data and actually generated power generation are matched, and estimates and outputs power generation from input data that is currently applied according to the learned method. The method of claim 1, wherein the time variable applying unit
Create two date variables with different offsets that increase or decrease continuously in a cycle of one year and two time variables with different offsets that increase or decrease in a continuous cycle of one day and increase or decrease continuously in a cycle of one day. Wind power generation prediction device for matching the time variable value to the weather information. The method of claim 1, wherein the atmospheric model application unit
an atmospheric stability calculation unit that calculates a bulk Richardson number (Bi _b ), which is an indicator of atmospheric dynamic stability, and calculates atmospheric stability (z _r /L) using the bulk Richardson number (Bi _b );
A zero-plane displacement calculation unit that calculates a stability variable (β) and a zero-plane displacement height (d _t ), which is a ground height newly defined in the atmospheric model, based on a canopy height (h), which is a height of an obstacle on the ground obtained by measurement;
a roughness length calculation unit that calculates a roughness length (z ₀ ) representing a height that becomes 0 when it is assumed that the wind follows a logarithmic profile; and
The friction velocity (u _* ) is calculated using the calculated atmospheric stability (z _r /L) and the roughness length (z ₀ ), and the wind speed (C _* ) above the canopy height (h) and below the canopy height (h) A wind power generation estimation device including a vertical wind speed determiner for obtaining a vertical wind speed profile by wind speed (C _* ) in. The method of claim 3, wherein the atmospheric stability calculator
As an indicator of the dynamic stability of the atmosphere, the bulk Richardson number (Bi _b ) used to determine whether turbulence occurs is calculated by the equation

(Where g is the gravitational acceleration, and θ _a is the potential temperature at the height (z _r ) of the lowest model layer (here, RSL) in the coordinate system of the existing WRF model (MM5 SL) of the numerical model) , where θ _va and θ _vg represent the virtual potential temperature at the SL and ground surface, respectively, and u(z _r ) represents the wind speed at the lowest model layer height (z _r ), and z represents the conventional It is a vertical coordinate value whose origin is the zero plane displacement height (d ₀ ) defined in the WRF model (MM5 SL).)
calculated according to
The atmospheric stability (z _r /L) is expressed by the equation

where u _* ^n-1 represents the friction velocity obtained in the previous step (n-1), ψ _m and ψ _h represent the integrated similarity function for momentum and heat, respectively,

,

denotes the integrated RSL function for momentum and heat, respectively. And ρ is the air density, c _p is the specific heat for air, and c _s represents the effective heat transfer coefficient. z ₀ and z _l are the roughness length and the viscous sublayer depth, respectively, and k is the von Karman constant (here 0.4).
wind power generation forecasting device that calculates according to The method of claim 4, wherein the zero plane displacement calculator
The zero plane displacement height (d _t ) is expressed by the equation

(Where l _m is the mixing length for momentum, β is the stability parameter, and L _c is the canopy penetration depth.)

(Where c _d is the drag coefficient at the leaf scale, a is the leaf area density, and LAI is the leaf area index.)

(Here, Φ _m is a similar function for momentum, and β _N represents the stability value in the natural state (=0.374).)
wind power generation forecasting device that calculates according to The method of claim 5, wherein the roughness length calculator
The roughness length (z ₀ ) is calculated by the equation

(here

represents the RSL similar function for momentum.)
wind power generation forecasting device that calculates according to The method of claim 6, wherein the vertical wind speed determiner
The frictional speed (u _* ) is calculated by the equation

calculated according to
The wind speed (C) above the canopy height (h) is expressed by the equation

(Where C _* is the scale value of the wind speed (C), C ₀ is the value of the variable (c) in the roughness length (z ₀ ),

is the integrated similarity function of the wind speed (C),

is the RSL function of the variable (C).)
calculated according to
The wind speed (C) below the canopy height (h) is expressed as

(Where C _h is the value of variable (c) at canopy height (h), and f is a parameter related to the depth scale of the scalar profile.)
A wind power generation forecasting device that obtains a vertical wind speed profile according to ground altitude by calculating according to. The method of claim 3, wherein the atmospheric model application unit
A standard deviation for a subgrid-scale orography shape of a predetermined size based on the location where the wind turbine is installed is obtained, and four neighboring subgrid-scale orography shapes located in the horizontal and vertical directions centered on a specific sub-grid are obtained. The Laplacian value of topographical altitude is obtained based on the height relationship with the grid, and different wind dissipation levels are calculated according to the standard deviation of the mountain topography on a sub-grid scale and the Laplacian value of topographical elevation according to the relationship with the surrounding topography. Wind power generation forecasting device further comprising a terrain-based wind speed correction unit that additionally reflects the vertical wind speed profile obtained by the vertical wind speed determination unit. The method of claim 1, wherein the weather information collecting unit
Based on the location where the wind power generator is installed, a forecast collection unit for collecting a forecast report predicting a meteorological condition at a predetermined time interval thereafter;
A wind power generation forecasting device comprising a nacelle data collection unit for collecting real-time weather information measured using a measuring device installed in the nacelle of the wind power generator. The method of claim 9, wherein the nacelle data collection unit
In order to obtain the learning data, the wind power generation amount predicting device for additionally collecting the amount of power currently produced by the wind power generator. Collecting weather information including weather forecasts;
Applying an atmospheric model that numerically analyzes the terrain altitude and surrounding environmental factors according to the installation location where the wind power generator is installed by receiving the weather forecast of low resolution in time and space to the turbulent spread to obtain high resolution;
Generating a plurality of time variables that continuously increase or decrease at predetermined intervals, and generating input data by matching time variable values according to the generated plurality of time variables with weather information including a plurality of high-resolution weather forecasts. ; and
A method for predicting wind power generation including the step of pre-learning based on learning data in which previously obtained input data and actual generated power are matched, and estimating and outputting the power generation from the currently applied input data according to the learned method. 12. The method of claim 11, wherein generating the input data comprises:
generating two date variables with different offsets that continuously increase or decrease with a cycle of one year and two time variables with different offsets that continuously increase or decrease with a cycle of one day; and
A method for predicting the amount of wind power generation comprising the step of matching a time variable value for a time to predict the amount of power generation with the weather information. The method of claim 11, wherein the step of high resolution
Calculating a bulk Richardson number (Bi _b ), which is an indicator of atmospheric dynamic stability, and calculating atmospheric stability (z _r /L) using the bulk Richardson number (Bi _b );
Calculating a stability variable (β) and calculating a zero plane displacement height (d _t ), which is a ground height newly defined in the atmospheric model, based on a canopy height (h), which is a height of an obstacle on the ground obtained by measurement;
Calculating a roughness length (z ₀ ) indicating a height at which the wind becomes 0 when it is assumed that the wind follows a logarithmic profile; and
The friction velocity (u _* ) is calculated using the calculated atmospheric stability (z _r /L) and the roughness length (z ₀ ), and the wind speed (C _* ) above the canopy height (h) and below the canopy height (h) A wind power generation prediction method comprising the step of obtaining a vertical wind speed profile by wind speed (C _* ) at . 14. The method of claim 13, wherein the step of calculating the atmospheric stability (z _r /L)
As an indicator of the dynamic stability of the atmosphere, the bulk Richardson number (Bi _b ) used to determine whether turbulence occurs is calculated by the equation

(Where g is the gravitational acceleration, and θ _a is the potential temperature at the height (z _r ) of the lowest model layer (here, RSL) in the coordinate system of the existing WRF model (MM5 SL) of the numerical model) , where θ _va and θ _vg represent the virtual potential temperature at the SL and ground surface, respectively, and u(z _r ) represents the wind speed at the lowest model layer height (z _r ), and z represents the conventional It is a vertical coordinate value whose origin is the zero plane displacement height (d ₀ ) defined in the WRF model (MM5 SL).)
calculating according to; and
The atmospheric stability (z _r /L) is expressed by the equation

where u _* ^n-1 represents the friction velocity obtained in the previous step (n-1), ψ _m and ψ _h represent the integrated similarity function for momentum and heat, respectively,

,

denotes the integrated RSL function for momentum and heat, respectively. And ρ is the air density, c _p is the specific heat for air, and c _s represents the effective heat transfer coefficient. z ₀ and z _l are the roughness length and the viscous sublayer depth, respectively, and k is the von Kårmån constant (here 0.4).
Wind power generation prediction method comprising the step of calculating according to. 15. The method of claim 14, wherein the step of calculating the zero plane displacement height (d _t )
The zero plane displacement height (d _t ) is expressed by the equation

(Where l _m is the mixing length for momentum, β is the stability parameter, and L _c is the canopy penetration depth.)

(Where c _d is the drag coefficient at the leaf scale, a is the leaf area density, and LAI is the leaf area index.)

(Here, Φ _m is a similar function for momentum, and β _N represents the stability value in the natural state (=0.374).)
Wind power generation forecasting method to calculate according to. 16. The method of claim 15, wherein calculating the roughness length (z ₀ )
The roughness length (z ₀ ) is calculated by the equation

(here

represents the RSL similar function for momentum.)
Wind power generation forecasting method to calculate according to. 17. The method of claim 16, wherein obtaining the vertical wind speed profile comprises:
The frictional speed (u _* ) is calculated by the equation

calculating according to;
The wind speed (C) above the canopy height (h) is expressed by the equation

(Where C _* is the scale value of the wind speed (C), C ₀ is the value of the variable (c) in the roughness length (z ₀ ),

is the integrated similarity function of the wind speed (C),

is the RSL function of the variable (C).)
calculating according to; and
The wind speed (C) below the canopy height (h) is expressed as

(Where C _h is the value of variable (c) at canopy height (h), and f is a parameter related to the depth scale of the scalar profile.)
Wind power generation prediction method comprising the step of calculating according to. The method of claim 13, wherein the high-resolution step
obtaining a standard deviation for a subgrid-scale orography shape having a predetermined size based on a location where the wind turbine is installed;
Obtaining a Laplacian value of topographic elevation based on a height relationship with four neighboring sub-grids located in horizontal and vertical directions around a specific sub-grid;
A wind power generation prediction method further comprising the step of additionally reflecting different wind dissipation levels to the vertical wind speed profile according to the Laplacian value of topographical altitude according to the relationship between the standard deviation of the mountain topography on a sub-grid scale and the surrounding topography. . 12. The method of claim 11, wherein collecting weather information comprises:
Collecting a forecast report predicting a weather condition at a predetermined time interval thereafter based on the location where the wind power generator is installed; and
Wind power generation prediction method comprising the step of collecting real-time meteorological information measured using a measurement method installed in the nacelle of the wind power generator. The method of claim 11, wherein the method for predicting the amount of wind power generation
To obtain the learning data, the wind power generation amount prediction method of additionally collecting the amount of power currently produced by the wind power generator.

Images