JP7686584B2 - Wind power generation forecasting system, wind power generation forecasting method, and program - Google Patents

Description

Embodiments of the present invention relate to a wind power generation forecasting system, a wind power generation forecasting method, and a program.

Conventionally, systems are known that predict weather using weather forecast data such as wind conditions. The weather forecast data calculated by this system can be used in a variety of situations that are affected by weather. For example, the weather forecast data can be used when predicting the amount of power generated by a wind power plant.

Patent No. 6326550

With the full liberalization of retail electricity sales, it is important for power generation companies and retail companies that handle renewable energy such as wind power to improve the accuracy of their power generation forecasts, both for the viability of their business and for grid stabilization.

However, weather forecasting models such as the Weather Research and Forecasting model (WRF model) used to forecast wind power generation generally predict wind speeds over a long period of time. As a result, they are not suitable for short-term electricity sales, and frequent revisions to power generation plans may be required.

The problem that the present invention aims to solve is to provide a wind power generation forecasting system, a wind power generation forecasting method, and a program that can improve the accuracy of short-term wind power generation forecasts.

A wind power generation prediction system according to one embodiment includes a wind condition observation unit that measures wind conditions in a wind farm for which wind power generation is to be predicted, a wind farm monitoring and control unit that records measurement data measured at the installation positions of the wind turbines in the wind farm for which wind power generation is to be predicted, a weather forecast data acquisition unit that acquires weather forecast data for an area including the wind farm for which wind power generation is to be predicted, a memory unit that stores wind condition observation data indicating the results of wind condition measurement, the measurement data, and weather analysis data indicating the results of analysis of the weather forecast data, and a prediction unit that uses the data stored in the memory unit to predict wind conditions for a first period including a prediction period for wind power generation and wind conditions for a second period including the prediction period but shorter than the first period, and predicts wind power generation based on the results of the wind condition prediction for the first period and the second period.

This embodiment makes it possible to improve the accuracy of forecasting short-term wind power generation.

1 is a block diagram showing a configuration of a wind power generation prediction system according to an embodiment. FIG. 2 is a diagram showing an example of the structure of a wind condition observation database. FIG. 2 is a diagram showing an example of a structure of a SCADA database. FIG. 2 is a diagram showing a part of the structure of a meteorological analysis database. FIG. 2 is a block diagram showing a configuration of a prediction unit. 4 is a flowchart showing an operation procedure for wind power generation prediction of the wind power generation prediction system. (a) is a diagram showing a schematic representation of the spatial resolution of weather forecast data, (b) is an example of weather forecast data, (c) is a diagram showing a schematic representation of the spatial resolution of meteorological analysis data, and (d) is an example of meteorological analysis data. 11 is a schematic diagram for explaining an example of a long-term prediction method of the long-term prediction section; FIG. 11 is a schematic diagram for explaining an example of a short-term prediction method of the short-term prediction section. FIG. 4 is a schematic diagram for explaining an example of a power generation amount prediction method of a power generation amount prediction unit. FIG. FIG. 13 is a block diagram showing a configuration of a prediction unit according to a first modified example. 13 is a graph showing an example of a comparison between predicted and measured wind speeds in the first modified example. FIG. 13 is a block diagram showing a configuration of a prediction unit according to a second modified example. FIG. 13 is a diagram illustrating an arrangement of wind turbines in a wind farm according to a second modified example. FIG. 13 is a diagram showing an example of changes in wind speed of an upstream wind turbine and a wake wind turbine in a second modified example.

The following describes an embodiment of the present invention with reference to the drawings. The following embodiment does not limit the present invention.

Figure 1 is a block diagram showing the configuration of a wind power generation prediction system according to one embodiment. The wind power generation prediction system 1 shown in Figure 1 includes a wind condition observation unit 100, a wind farm monitoring control unit 200, a weather forecast data acquisition unit 300, a data analysis processing unit 500, a memory unit 600, and a prediction unit 700. In the wind power generation prediction system 1, the wind condition observation unit 100, the wind farm monitoring control unit 200, and the weather forecast data acquisition unit 300 are each connected to the data analysis processing unit 500 via a communication network 400.

In addition, in the wind power generation prediction system 1, the data analysis processing unit 500, the memory unit 600, and the prediction unit 700 may be provided as independent devices, or each unit may be provided in one server device. Also, the data analysis processing unit 500 and the prediction unit 700 may be provided in this server device, and the memory unit 600 may be provided as a device independent of the server device.

The wind condition observation unit 100 includes a measurement unit 110 and a measurement control unit 120. The measurement unit 110 measures wind conditions, for example, by scanning LiDAR (Light Detection And Ranging). LiDAR is a measuring instrument that emits laser light into the atmosphere, receives scattered light from the atmosphere, and observes wind speed and direction from the Doppler frequency. This wind condition measurement does not have to be performed by scanning LiDAR, and any means for measuring wind conditions in the sky, such as an observation mast or vertical LiDAR, will do. At least one measurement unit 110 is installed at any position within the wind farm for which wind power generation is to be predicted or within the surrounding area.

The measurement control unit 120 sets the measurement conditions of the measurement unit 110. The measurement conditions include, for example, the measurement range, the measurement position, and the measurement frequency. It is desirable that the measurement conditions can be changed remotely, for example, via the communication unit 510 of the data analysis processing unit 500, but this is not necessarily the case. In addition, the measurement control unit 120 may extract wind direction information from the measurement data of the measurement unit 110 and change the measurement position of the measurement unit 110 so as to measure the wind conditions at a wind turbine located upstream (upwind) in the wind farm, or at a position in front of the turbine.

The wind farm monitoring control unit 200 acquires supervisory control and data acquisition (SCADA) data for each wind turbine installed in the wind farm. The SCADA data includes measurement data such as wind speed, wind direction, and the amount of power generated by the wind turbine at the installation location of the wind turbine.

The weather forecast data acquisition unit 300 acquires, for example, weather forecast GPV (Grid Point Value) data, which is weather forecast data distributed by the Japan Meteorological Agency. Weather forecast GPV data is past and future weather forecast data calculated by a supercomputer at grid points preset on a map. Note that the weather forecast data acquired by the weather forecast data acquisition unit 300 does not necessarily have to be of one type.

The data analysis processing unit 500 has a communication unit 510, a data processing unit 520, and a weather forecast data analysis unit 530. Each unit of the data analysis processing unit 500 is described below.

The communication unit 510 functions as a communication interface when performing data communication with the wind condition observation unit 100, the wind farm monitoring control unit 200, and the weather forecast data acquisition unit 300 via the communication network 400. Furthermore, the communication unit 510 transfers data acquired from the wind condition observation unit 100 and the wind farm monitoring control unit 200 to the data processing unit 520, and transfers weather forecast data acquired from the weather forecast data acquisition unit 300 to the weather forecast data analysis unit 530.

The data processing unit 520 performs primary processing of the wind observation data measured by the wind observation unit 100. Primary processing of the wind observation data includes, for example, performing vector synthesis, vector averaging, and other processing on the line-of-sight wind speed measured by the wind observation unit 100, and calculating the horizontal and vertical wind speed, wind direction, turbulence intensity, wind direction standard deviation, and the like at the measurement position.

The data processing unit 520 also performs primary processing of the SCADA data measured by the wind farm monitoring control unit 200. Primary processing of the SCADA data includes, for example, classifying the SCADA data by referring to the operating status of the wind turbines, and calculating the wind speed ratio between the wind speed of the wind turbine located upstream and the wind speed of the wind turbine located downstream in the wind farm by referring to wind direction information.

The weather forecast data analysis unit 530 has a function of outputting weather analysis data that indicates the analysis results obtained by adjusting the time resolution and spatial resolution of the weather forecast data acquired by the weather forecast data acquisition unit 300. For example, the weather forecast data analysis unit 530 can perform nesting using a WRF analysis model such as a numerical weather model, thereby increasing the spatial resolution and time resolution of the weather analysis data compared to the weather forecast data.

The memory unit 600 has a wind condition observation database 610, a SCADA database 620, and a weather analysis database 630. Each database is described below.

Figure 2 is a diagram showing an example of the structure of the wind observation database 610. The wind observation database 610 stores the results of the data processing unit 520 performing primary processing on the wind observation data. The wind observation database 610 shown in Figure 2 stores the time of measurement by the wind observation unit 100, the latitude, longitude, and altitude of the measurement point, the horizontal wind speed, the horizontal wind direction, turbulence intensity, etc.

Figure 3 is a diagram showing an example of the structure of the SCADA database 620. The SCADA database 620 stores the results of the data processing unit 520 performing primary processing of the SCADA data. The SCADA database 620 shown in Figure 3 stores the time measured by the wind farm monitoring control unit 200, the wind turbine number of the monitored target, horizontal wind speed, horizontal wind direction, various data related to the wind turbine such as power generation amount, and the wake state and inflow wind speed ratio calculated by the data processing unit 520.

Figure 4 shows a portion of the structure of the weather analysis database 630. The weather analysis database 630 stores weather analysis data that indicates the results of the weather forecast data analysis unit 530 analyzing the weather forecast data. The weather analysis database 630 shown in Figure 4 stores the forecast time, analysis point, altitude of the analysis point, north-south wind, east-west wind, air pressure, etc., indicated in the weather forecast data.

Figure 5 is a block diagram showing the configuration of the prediction unit 700. The prediction unit 700 has a short-term prediction unit 710, a long-term prediction unit 720, a power generation prediction unit 730, and a result display unit 740. Each part of the prediction unit 700 will be described below.

The short-term prediction unit 710 has a function of predicting future short-term wind speeds using a short-term prediction model created based on data stored in the wind condition observation database 610 or the SCADA database 620. Specifically, the short-term prediction unit 710 has a short-term data preprocessing unit 711, a short-term wind condition learning unit 712, and a short-term calculation unit 713.

The short-term data preprocessing unit 711 processes the wind observation data read from the wind observation database 610 or the SCADA data read from the SCADA database 620 to create teacher data for machine learning. The short-term wind condition learning unit 712 performs machine learning on the teacher data of the short-term data preprocessing unit 711 to create a short-term prediction model. The short-term calculation unit 713 performs calculations on the latest wind observation data read from the wind observation database 610 or the latest SCADA data read from the SCADA database 620 using the short-term prediction model. This allows the short-term predicted wind speed, including the prediction period for wind power generation, to be calculated.

The long-term prediction unit 720 has a function of correcting the meteorological analysis data using a long-term prediction model created based on the SCADA data stored in the SCADA database 620 and the meteorological analysis data stored in the meteorological analysis database 630. Specifically, the long-term prediction unit 720 has a long-term data preprocessing unit 721, a long-term wind condition learning unit 722, and a long-term calculation unit 723.

The long-term data pre-processing unit 721 processes the SCADA data read from the SCADA database 620 and the meteorological analysis data read from the meteorological analysis database 630 to create teacher data for machine learning. The long-term wind condition learning unit 722 performs machine learning on the teacher data from the long-term data pre-processing unit 721 to create a long-term prediction model. The long-term calculation unit 723 performs calculations on the latest meteorological analysis data read from the meteorological analysis database 630 using the long-term prediction model. This allows correction data for the long-term meteorological analysis data, including the forecast period for wind power generation, to be calculated.

The power generation prediction unit 730 has a function of predicting the amount of wind power generation using a generator model created based on the SCADA data stored in the SCADA database 620. Specifically, the power generation prediction unit 730 has a generator data preprocessing unit 731, a power generation learning unit 732, and a power generation calculation unit 733.

The generator data preprocessing unit 731 processes the SCADA data read from the SCADA database 620 to create teacher data for machine learning. The power generation amount learning unit 732 uses the teacher data created by the generator data preprocessing unit 731 in the neural network NN2 to create a generator model for predicting wind power generation. The power generation amount calculation unit 733 performs calculations using the short-term predicted wind speed calculated by the short-term prediction unit 710 and the correction data calculated by the long-term prediction unit 720, using the generator model created by the power generation amount learning unit 732. This calculates the wind power generation amount for the prediction period.

The result display unit 740 displays various images such as the wind power generation amount calculated by the power generation amount prediction unit 730. The result display unit 740 has a display device such as a liquid crystal display. In this embodiment, the result display unit 740 is configured as part of the prediction unit 700, but may be configured independent of the prediction unit 700.

Next, a method for predicting wind power output using the wind power output prediction system 1 according to this embodiment will be described with reference to FIG. 6. FIG. 6 is a flowchart showing the operational procedure for predicting wind power output by the wind power output prediction system 1 according to this embodiment. The operations performed according to the steps of this flowchart can also be realized by having a computer execute a program that processes each step. This program can also be recorded on a recording medium as software.

In the flowchart shown in FIG. 6, first, in the data analysis processing unit 500, the communication unit 510 acquires wind observation data from the wind observation unit 100, SCADA data from the wind farm monitoring control unit 200, and weather forecast data from the weather forecast data acquisition unit 300 via the communication network 400 (step S11).

Next, in the data analysis processing unit 500, the data processing unit 520 performs primary processing of the wind observation data and the SCADA data, and the weather forecast data analysis unit 530 analyzes the weather forecast data (step S12). Here, an example of a weather analysis method of the weather forecast data analysis unit 530 will be described with reference to Figures 7(a) to 7(d).

Figure 7(a) is a diagram showing a schematic representation of the spatial resolution of weather forecast data, and Figure 7(b) is an example of weather forecast data. Figure 7(c) is a diagram showing a schematic representation of the spatial resolution of weather analysis data, and Figure 7(d) is an example of weather analysis data.

The weather forecast data to be analyzed by the weather forecast data analysis unit 530 is created in units of grids when the analysis target area including the installation area of the wind farm to be predicted for wind power generation and its surrounding area is divided into a grid, as shown in FIG. 7(a). Note that the prediction target wind farm shown in FIG. 7(a) is an offshore wind farm, but it may also be an onshore wind farm.

Figure 7(b) shows weather forecast data for grid point A shown in Figure 7(a). This weather forecast data shows, for example, north-south wind speed, east-west wind speed, temperature, and changes in temperature over time at grid point A. Note that while Figure 7(b) shows weather forecast data for one altitude at grid point A, the weather forecast data for grid point A is created for multiple altitudes that are set in stages in advance. The data to be analyzed includes not only the weather forecast data for grid point A, but also weather forecast data for other grid points.

As shown in FIG. 7(c), the weather forecast data analysis unit 530 divides the analysis area into a grid smaller than the weather forecast data according to the required forecast accuracy. This results in a higher spatial resolution than the weather forecast data. The weather forecast data analysis unit 530 creates weather analysis data for each grid point by performing nesting using a WRF analysis model such as a numerical weather model with the weather forecast data as input data.

Figure 7(d) shows meteorological analysis data at one altitude at grid point a. This meteorological analysis data shows, for example, north-south wind speed, east-west wind speed, temperature, and changes in temperature over time at grid point a. At this time, while the weather forecast data shows changes over time in one-hour increments, the meteorological analysis data shows changes over time in ten-minute increments. Therefore, the time resolution of the meteorological analysis data is higher than that of the meteorological forecast data. Note that while Figure 7(d) shows meteorological analysis data at one altitude at grid point a, the meteorological analysis data at grid point a is created for multiple altitudes, just like the weather forecast data. In addition, the meteorological analysis data includes data not only for grid point a, but also for other grid points.

As described above, the weather analysis data created by the weather forecast data analysis unit 530 is stored in the weather analysis database 630 (step S13). Also, in step S13, the wind observation data and SCADA data that have been subjected to primary processing by the data processing unit 520 are stored in the wind observation database 610 and the SCADA database 620, respectively.

Next, the long-term prediction unit 720 of the prediction unit 700 performs a long-term prediction of the wind speed (step S14). Here, the operation of step S14 will be described with reference to FIG. 8.

Figure 8 is a schematic diagram for explaining an example of the long-term prediction method of the long-term prediction unit 720. In step S14, first, the long-term data pre-processing unit 721 of the long-term prediction unit 720 reads out past SCADA data from the SCADA database 620 and reads out past meteorological analysis data from the meteorological analysis database 630. The past SCADA data and past meteorological analysis data read out by the long-term data pre-processing unit 721 include, for example, past data for the same time as the predicted time of wind power generation. Next, the long-term data pre-processing unit 721 processes the read SCADA data and meteorological analysis data to create teacher data for machine learning.

Next, the long-term wind condition learning unit 722 creates a long-term prediction model based on machine learning using the teacher data created by the long-term data preprocessing unit 721. In this embodiment, this long-term prediction model is a mathematical model that can be used in the neural network NN1.

Finally, the long-term calculation unit 723 reads the latest meteorological analysis data from the meteorological analysis database 630. Next, the long-term calculation unit 723 performs calculations using the read meteorological analysis data as input data with the above-mentioned long-term prediction model. As a result, correction data for the meteorological analysis data is calculated. This correction data is wind speed correction data at the installation site of the wind turbine that is the subject of the wind power generation prediction.

The wind speed correction data is calculated using a long-term prediction model generated by machine learning using past data as training data. Therefore, as shown in Figure 8, the error of the wind speed correction data with respect to the actual wind speed measurement data is smaller than that of the meteorological analysis data. Note that the method of calculating the correction data is not limited to the above method.

When the long-term prediction unit 720 finishes the long-term prediction as described above, the short-term prediction unit 710 of the prediction unit 700 then performs a short-term prediction of the wind speed (step S15). Here, the operation of step S15 will be described with reference to FIG. 9.

Figure 9 is a schematic diagram for explaining an example of the short-term prediction method of the short-term prediction unit 710. In step S15, first, the short-term data pre-processing unit 711 of the short-term prediction unit 710 reads out past SCADA data from the SCADA database 620, or reads out past wind condition observation data from the wind condition observation database 610. The past SCADA data or past wind condition observation data read out by the short-term data pre-processing unit 711 includes, for example, past data for the same time as the predicted time of wind power generation.

Whether the short-term data pre-processing unit 711 reads out SCADA data or wind observation data is determined in advance according to the contents of the short-term prediction model described below. When a measurement item that is not in the SCADA data but is in the wind observation data, such as wind speed standard deviation, is required to create a short-term prediction model, the short-term data pre-processing unit 711 reads out the wind observation data. In this case, the short-term data pre-processing unit 711 processes the read out wind observation data to create teacher data for machine learning. When the short-term data pre-processing unit 711 reads out SCADA data, it processes the SCADA data to create teacher data for machine learning.

Next, the short-term wind condition learning unit 712 creates a short-term prediction model based on machine learning using the teacher data created by the short-term data preprocessing unit 711. In this embodiment, this short-term prediction model is a mathematical model that can be used in the recurrent neural network RNN1.

Finally, the short-term calculation unit 713 reads the latest SCADA data from the SCADA database 620, or reads the latest wind observation data from the wind observation database 610. Next, the short-term calculation unit 713 performs calculations using the read SCADA data or wind observation data as input data with the above-mentioned short-term prediction model. As a result, the short-term predicted wind speed is calculated. When the input data is SCADA data, this short-term predicted wind speed is wind speed data at the installation site of the wind turbine that is the subject of the wind power generation prediction. On the other hand, when the input data is wind observation data, this short-term predicted wind speed is wind speed data at the wind observation site within the subject area of the wind power generation prediction.

The above short-term predicted wind speed is calculated using a short-term prediction model generated by machine learning using past actual measurement data as training data. In other words, the short-term prediction model is created using only actual measurement data without using prediction data. Therefore, it is possible to calculate the short-term wind speed at the predicted point of wind power generation with high accuracy. Note that the method of calculating the short-term predicted wind speed is not limited to the above method.

When the short-term prediction unit 710 finishes the short-term prediction as described above, the power generation amount prediction unit 730 of the prediction unit 700 then predicts the wind power generation amount (step S16). Here, the operation of step S16 will be described with reference to FIG. 10.

Figure 10 is a schematic diagram for explaining an example of a power generation prediction method of the power generation prediction unit 730. In step S15, first, the generator data pre-processing unit 731 of the power generation prediction unit 730 reads past SCADA data from the SCADA database 620. The past SCADA data read by the generator data pre-processing unit 731 includes, for example, past data for the same time as the predicted time of wind power generation. Next, the generator data pre-processing unit 731 processes the read SCADA data to create teacher data for machine learning.

Next, the power generation amount learning unit 732 creates a generator model based on machine learning using the teacher data created by the generator data preprocessing unit 731. In this embodiment, this generator model is a mathematical model that can be used in the neural network NN2.

Finally, the power generation amount calculation unit 733 performs calculations using the short-term predicted wind speed calculated by the short-term prediction unit 710 and the correction data calculated by the long-term prediction unit 720 as input data in the above-mentioned generator model. As a result, a predicted power generation amount is calculated. This predicted power generation amount is displayed on the result display unit 740 as a graph showing the relationship between wind speed and power generation amount based on the SCADA data, as shown in FIG. 10, for example.

In a wind farm, the amount of wind power generated depends on the wind speed. Wind speed not only changes over long periods of several hours, but can also change over short periods of several seconds. Therefore, when predicting short-term wind power generation, the accuracy is insufficient if the prediction is based solely on short-term wind speed predictions.

In this embodiment, the long-term prediction unit 720 predicts long-term (first period) wind conditions including the prediction period for wind power generation, and the short-term prediction unit 710 predicts short-term (second period) wind conditions including the prediction period. Furthermore, the power generation amount calculation unit 733 calculates the wind power generation amount for the prediction period using the prediction results of the long-term prediction unit 720 and the short-term prediction unit 710. In this way, by predicting the wind power generation amount using not only the short-term predicted wind speed but also the long-term predicted wind speed, it is possible to predict the short-term wind power generation amount with high accuracy.

(First Modification)
Fig. 11 is a block diagram showing the configuration of a prediction unit 700 according to the first modified example. In Fig. 11, the same components as those in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

In this modified example, the short-term prediction unit 710 calculates the short-term predicted wind speed using the correction data that is the prediction result of the long-term prediction unit 720. Here, the short-term wind speed prediction method according to this modified example will be described with reference to FIG. 12.

Figure 12 is a graph showing an example of a comparison between the predicted wind speed and the actual measurement. In Figure 12, the horizontal axis indicates time, and the vertical axis indicates wind speed. The solid line indicates the prediction result of the short-term prediction unit 710, the dashed line indicates the prediction result of the long-term prediction unit 720, and the dotted line indicates the actual measurement result. In Figure 12, at time t0, the wind speed predicted by the long-term prediction unit 720 changes significantly. If the change in wind speed at this time exceeds a preset reference value, the short-term prediction unit 710 calculates a short-term predicted wind speed using the prediction result of the long-term prediction unit 720. Specifically, when the short-term wind condition learning unit 712 of the short-term prediction unit 710 creates a short-term prediction model, it sets a weighting coefficient for the short-term prediction model according to the prediction result of the long-term prediction unit 720.

When the short-term wind condition learning unit 712 creates a short-term prediction model as described above, the short-term calculation unit 713 performs calculation processing using the short-term prediction model to calculate the short-term predicted wind speed. After that, the power generation amount prediction unit 730 predicts the wind power generation amount based on the short-term predicted wind speed. Finally, the result display unit 740 displays the prediction result of the power generation amount prediction unit 730.

According to the present modified example described above, the wind power generation amount is predicted by reflecting the prediction result of the long-term prediction unit 720 in the prediction result of the short-term prediction unit 710. Therefore, in this modified example as well, the wind power generation amount is predicted using the prediction result of the long-term prediction unit 720 and the prediction result of the short-term prediction unit 710. Therefore, it is possible to predict the short-term wind power generation amount with high accuracy.

(Second Modification)
Fig. 13 is a block diagram showing the configuration of a prediction unit 700 according to the second modified example. In Fig. 13, the same components as those in the above-described embodiment are given the same reference numerals, and detailed description thereof will be omitted.

In this modified example, the prediction unit 700 further includes a wind speed correction unit 750. The wind speed correction unit 750 corrects the wind speed of the wake wind turbine in the wind farm using a wake model. The wake model will now be described with reference to Figures 14 and 15.

Figure 14 is a diagram showing the arrangement of wind turbines in a wind farm. Also, Figure 15 is a diagram showing an example of the change in wind speed of an upstream wind turbine and a wake wind turbine.

As shown in FIG. 14, for example, if the wind direction is westward, going from left to right on the drawing, the upstream wind turbine is a wind turbine located on the westward, upwind side, and the wake wind turbine is a wind turbine located on the eastward, downwind side. In this case, as shown in FIG. 15, in the time period from time t1 to time t2, the wind speed of the wake wind turbine slows down due to the influence of the upstream wind turbine, and in the time period from time t2 to time t3, the wind speed of the wake wind turbine is approximately equal to the wind speed of the upstream wind turbine.

The wind speed deceleration of the wake wind turbine caused by the upstream wind turbine affects the amount of power generated by the wake wind turbine. The wake model is a mathematical model for calculating the degree of impact that the wind speed deceleration of the wake wind turbine caused by the upstream wind turbine has on the amount of power generated by the wake wind turbine. The wind speed correction unit 750 uses the wind direction data included in the wind conditions predicted by the short-term prediction unit 710 and the long-term prediction unit 720 to identify the wind turbine installation positions within the wind farm and the wake wind turbines that are affected by the wind speed deceleration caused by the upstream wind turbine based on their positional relationship. The wind speed correction unit 750 then corrects the wind speed of the wake wind turbine using the wake model.

Therefore, according to this modified example, the wind speed of the wake wind turbine can be corrected from the wind speed of the upstream wind turbine predicted by the short-term prediction unit 710 and the long-term prediction unit 720, respectively, improving the accuracy of the wind speed prediction. As a result, the accuracy of the wind power generation amount predicted based on the corrected wind speed of the wake wind turbine is also improved.

Although the above describes the embodiments and modifications, these embodiments and modifications are presented only as examples and are not intended to limit the scope of the invention. The novel system described in this specification can be embodied in various other forms. Furthermore, various omissions, substitutions, and modifications can be made to the forms of the system described in this specification without departing from the spirit of the invention. The appended claims and equivalents are intended to include such forms and modifications that fall within the scope and spirit of the invention.

1: Wind power generation prediction system 100: Wind condition observation unit 200: Wind farm monitoring and control unit 300: Weather forecast data acquisition unit 530: Weather forecast data analysis unit 600: Memory unit 700: Prediction unit 710: Short-term prediction unit 711: Short-term data pre-processing unit 712: Short-term wind condition learning unit 713: Short-term calculation unit 720: Long-term prediction unit 721: Long-term data pre-processing unit 722: Long-term wind condition learning unit 723: Long-term calculation unit 730: Power generation prediction unit 731: Power generator data pre-processing unit 732: Power generation learning unit 733: Power generation calculation unit 750: Wind speed correction unit

Claims (10)

a wind condition observation unit that measures wind conditions in a wind farm for which wind power generation is to be predicted;
A wind farm monitoring control unit that records measurement data measured at installation positions of wind turbines in the wind farm to be predicted;
A weather forecast data acquisition unit that acquires weather forecast data in an area including the prediction target wind farm;
a storage unit for storing wind condition observation data indicating the measurement results of the wind conditions, the measurement data, and meteorological analysis data indicating the analysis results of the meteorological forecast data;
a prediction unit that predicts wind conditions for a first period including a prediction period of the wind power generation amount and for a second period including the prediction period and shorter than the first period using the data stored in the storage unit, and predicts the wind power generation amount based on the wind condition prediction results for the first period and the second period;
Equipped with
The prediction unit predicts wind conditions in the second time period using a wind condition prediction result in the first time period . The prediction unit includes a short-term prediction unit that predicts wind conditions for the second period,
The short-term prediction unit
a short-term data preprocessing unit that processes the wind observation data or the measurement data read from the storage unit to create teacher data for machine learning;
A short-term wind condition learning unit that creates a short-term prediction model based on machine learning using the teacher data;
The wind power generation prediction system according to claim 1 , further comprising: a short-term calculation unit that calculates the wind conditions in the second time period by a calculation process using the short-term prediction model. The wind power generation prediction system according to claim 2 , wherein the short-term calculation unit performs calculation processing based on the short-term prediction model within a recurrent neural network. The prediction unit includes a long-term prediction unit that predicts wind conditions for the first period,
The long-term prediction unit
a long-term data preprocessing unit that processes the measurement data and meteorological analysis data read from the storage unit to create teacher data for machine learning;
A long-term wind condition learning unit that creates a long-term prediction model based on machine learning using the teacher data;
The wind power generation prediction system according to claim 1 , further comprising: a long-term calculation unit that calculates the wind conditions for the first period by a calculation process using the long-term prediction model. The prediction unit includes a power generation amount prediction unit that predicts the wind power generation amount,
The power generation amount prediction unit
a generator data preprocessing unit that processes the measurement data read from the storage unit to create teacher data for machine learning;
a power generation amount learning unit that creates a power generator model based on machine learning using the teacher data;
The wind power generation prediction system according to claim 1 , further comprising: a power generation amount calculation unit that calculates the wind power generation amount by a calculation process using the generator model. The wind power generation prediction system according to claim 5 , wherein the power generation amount calculation unit performs calculation processing based on the generator model within a neural network. a weather forecast data analysis unit that analyzes the weather forecast data acquired from the weather forecast data acquisition unit and outputs the weather analysis data;
7. The wind power generation prediction system according to claim 1 , wherein the weather forecast data analysis unit adjusts a time resolution and a spatial resolution of the weather forecast data according to a required prediction accuracy and outputs the weather analysis data. In the prediction target wind farm, an upstream wind turbine arranged on the windward side and a wake wind turbine arranged on the leeward side are installed,
8. The wind power generation prediction system according to claim 1, wherein the prediction unit includes a wind speed correction unit that corrects the wind speed of the wake wind turbine using a wake model for calculating an effect of a wind speed deceleration of the wake wind turbine caused by the upstream wind turbine on an amount of power generation of the wake wind turbine. Measure the wind conditions at the wind farm for which wind power generation is to be predicted,
Recording measurement data measured at installation positions of wind turbines in the wind farm to be predicted;
Obtaining weather forecast data for an area including the prediction target wind farm;
storing wind condition observation data indicating the measurement results of the wind conditions, the measurement data, and meteorological analysis data indicating the analysis results of the meteorological forecast data in a storage unit;
Using the data stored in the storage unit, predicting wind conditions for a first period including a prediction period of the wind power generation amount and wind conditions for a second period including the prediction period and shorter than the first period, wherein the wind conditions for the second period are predicted using a wind condition prediction result for the first period;
predicting the wind power generation amount based on wind condition prediction results for the first period and the second period;
Wind power forecasting methods. Measure the wind conditions at the wind farm for which wind power generation is to be predicted,
Recording measurement data measured at installation positions of wind turbines in the wind farm to be predicted;
Obtaining weather forecast data for an area including the prediction target wind farm;
storing wind condition observation data indicating the measurement results of the wind conditions, the measurement data, and meteorological analysis data indicating the analysis results of the meteorological forecast data in a storage unit;
Using the data stored in the storage unit, predicting wind conditions for a first period including a prediction period of the wind power generation amount and wind conditions for a second period including the prediction period and shorter than the first period, wherein the wind conditions for the second period are predicted using a wind condition prediction result for the first period;
A program for causing a computer to execute a process of predicting the wind power generation amount based on wind condition prediction results for the first period and the second period.