Meteorological data-based provincial multi-year historical wind power generation output inversion method and device

Abstract

The invention provides a method and equipment for inverting provincial multi-year historical wind power generation output based on meteorological data, and the method comprises the steps: building a generation output regression model of each wind power plant according to the hourly wind speed and wind direction data of the 100m height of ERA5 meteorological reanalysis data of the actual position of each wind power plant in the provincial region; and then 100 m height wind speed and wind direction data of years of historical data are analyzed by adopting ERA5 meteorology of the position of each wind power plant, hourly power generation output data of the wind power plant under historical meteorological conditions are inversely generated, and the hourly power generation output data of years of historical wind power generation in the provincial region can be obtained by accumulating the historical power generation output of each wind power plant in the provincial region. According to the method, provincial wind power generation output data under historical meteorological conditions for many years before the wind power plant is built can be generated, and the understanding of provincial wind power generation output law characteristics under various meteorological scenes can be deepened.

Description

Technical Field

[0001] The invention belongs to the field of wind power generation, and in particular relates to a method and a device for inverting the historical wind power generation output of a province over many years based on meteorological data. Background Art

[0002] In recent years, the proportion of wind power in my country's power system power supply has increased rapidly. Affected by the inherent uncertainty and randomness of wind power, large-scale wind power access has brought prominent contradictions in the balance of power and electricity to the power system. Under normal circumstances, the capacity factor of the power output of multiple wind farms in a province is generally above 0.2. Under unfavorable meteorological conditions of large-scale calm wind, the provincial wind power output will have an extremely low power generation capacity factor because a large number of wind farms cannot generate electricity. When the proportion of wind power in the power supply is low, other power outputs can be adjusted to ensure power supply; however, in power grids with a high proportion of wind power access, large-scale power restrictions may occur due to insufficient power output. At present, the proportion of wind power in some areas of my country has reached 20%, and power restrictions caused by abnormal wind power output have become more and more prominent. For example, in September 2021, a large area of ​​calm wind lasted for a week in Northeast my country, and the local wind power output capacity factor fell from about 0.2 under normal conditions to below 0.1, which led to power restrictions for a large number of users. In order to ensure the power supply reliability of high-proportion wind power access to the power grid, it is urgent to study and master the fluctuation characteristics and general laws of large-scale wind power output in the province.

[0003] Traditionally, the power system mainly analyzes the power balance of power output and load demand based on typical daily data of power sources and power loads and the statistical characteristics of the whole year. Since the power output of hydropower, coal-fired power generation and nuclear power is basically stable and controllable, and the output of wind power generation depends on wind resource conditions, the power balance of the power system with a high proportion of wind power access is significantly affected by meteorological conditions. The power system accumulates operating experience based on a large amount of historical operating data. However, wind farm construction is currently in a period of rapid growth, with a development time of generally less than 10 years, and the accumulated operating data and experience are limited. At present, there is still a lack of in-depth understanding of how provincial wind power output will perform under various complex meteorological scenarios.

[0004] At present, there are studies on wind power output prediction. For example, patent application CN111062516 discloses a wind turbine output prediction method, which predicts wind turbine output by using regional meteorological data (wind speed data) at the time to be predicted. Conventional wind power output prediction is based on the output data of the current wind farm or the meteorological forecast data of the future location of the wind farm to predict future wind power output.

[0005] There are other studies based on ERA5 meteorological data to analyze wind power output in vast areas. These methods do not consider the actual location of wind farms, but are based on the wind resource conditions of each grid. It is believed that wind farms should be located more in areas with better wind resource conditions to set the weight of each grid. It is generally believed that grids with better wind resource conditions have more wind farms and higher weights, while grids with poorer wind resource conditions are set to have no wind farms and set the weight to zero. On this basis, the historical wind power output data of the province for many years is inverted. In fact, the site selection of wind farms does not only consider wind resource conditions, but also the environmental noise during the operation of wind turbines, grid access conditions, etc. Judging from the actual location of wind farms in provincial power grids, there is a large degree of deviation from wind resource conditions. Summary of the invention

[0006] The purpose of the present invention is to address the deficiencies in the prior art and to provide a method and device for inverting the historical wind power output of a province over many years based on meteorological data, so as to improve the accuracy of the inversion analysis of the historical wind power output of a province over many years.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A method for inverting provincial wind power output over many years based on meteorological data comprises the following steps:

[0009] S1: The actual locations of each wind farm in the province correspond to the cells of the ERA5 meteorological reanalysis database, and the hourly meridional wind speed and zonal wind speed at a height of 100m corresponding to the cells of the ERA5 meteorological reanalysis data are obtained;

[0010] S2: Based on the ERA5 meteorological reanalysis data, calculate the wind speed and wind direction of each cell where the wind farm is located, and the wind speed V of the i-th cell i The expression is as follows: The wind direction θ of the i-th cell i The expression is as follows:

[0011] S3: Input the historical hourly wind speed, wind direction and temperature data of the cells where each wind farm is located for many years into the pre-trained wind farm power output regression model, and output the historical hourly power output data of each wind farm for many years;

[0012] S4: adding the historical hourly power output data of each wind farm in the same period for many years obtained in S3 to obtain the historical hourly power output data of the province for many years;

[0013] Among them, u i is the meridional wind speed at a certain moment at a height of 100m in the i-th cell, v iis the latitudinal wind speed at the i-th cell at a height of 100 m at that moment, i = 1, 2…k, k is the number of wind farms in the area.

[0014] The present invention uses ERA5 data obtained from the grid corresponding to the longitude and latitude of the wind farm to invert historical wind power output; the present invention establishes a wind farm power output regression model of the actual wind farm based on the actual wind farm location in the province, which can effectively avoid the inversion deviation of wind power output data caused by the wind farm location deviation, and can better reveal the regular characteristics of provincial wind power output under complex meteorological scenes. The present invention uses the wind speed and wind direction data of the actual location of each wind farm to perform provincial wind power analysis, and the reconstructed provincial wind farm power output is more accurate.

[0015] Furthermore, the wind farm power generation output regression model includes an iTransformer neural network, an LSTM neural network, and a CNN neural network.

[0016] Furthermore, the training process of the wind farm power generation output regression model includes:

[0017] Obtain the hourly measured wind power output of the wind farm for one year and the wind speed, wind direction, and temperature of the cell where the wind farm is located;

[0018] 75% of the acquired data is used as a training set, and 25% of the data is used as a test set to train the wind power regression model of the wind farm;

[0019] The absolute error RMSE, mean square error NRMSE and determination coefficient R of the wind farm measured power output and the wind farm regression power output are calculated. 2 Evaluate the performance of wind farm wind power output regression models.

[0020] Furthermore, a regression model for the power generation output of all wind farms in the province is established one by one, and after generating the historical hourly power generation output data of all wind farms for many years, the historical wind power generation output data of the province for many years is accumulated.

[0021] Based on the same inventive concept, the present invention also provides an electronic device, including:

[0022] one or more processors;

[0023] A memory having one or more programs stored thereon, which, when executed by the one or more processors, enables the one or more processors to implement the steps of a method for inverting wind power output based on meteorological data.

[0024] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a method for inverting wind power output based on meteorological data.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention establishes a wind farm power output regression model of the actual wind farm based on the actual wind farm location in the province, which can effectively avoid the inversion deviation of wind power output data caused by the wind farm location deviation, and can better reveal the regular characteristics of provincial wind power output under complex meteorological scenes. The present invention uses the wind speed and wind direction data of the actual location of each wind farm to perform provincial wind power analysis, and the reconstructed provincial wind farm power output is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the iTransformer model structure of the present invention;

[0028] Figure 2 This is the result diagram of the iTransformer model of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be described in detail below in conjunction with the embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0030] Example

[0031] The method for inverting wind power output based on meteorological data in this embodiment includes the following process:

[0032] 1. Obtain the measured wind power of all wind farms in a certain area in a certain year, and obtain the meteorological data (u, v wind speed and temperature) at a height of 100m in the grid where the wind farm area is located in the ERA5 meteorological reanalysis database; divide the area into multiple cells according to the resolution of the ERA5 meteorological reanalysis database; ERA5 (ECMWF Reanalysis v5) is the fifth generation of atmospheric reanalysis data set of the global climate from January 1950 to the present by the European Center for Medium-Range Weather Forecasts, with fine data resolution (1hour, 0.25°*0.25°).

[0033] 2. Calculate the wind speed and direction of each cell where the wind farm is located, and the wind speed V of the ith cell i The expression is as follows: The wind direction θ of the i-th cell i The expression is as follows Among them, u i is the meridional wind speed at a certain moment at a height of 100m in the i-th cell, v i is the latitudinal wind speed at the i-th cell at a height of 100 m at that moment, i = 1, 2…k, k is the number of wind farms / cells in the region.

[0034] 3. Take the hourly wind speed, wind direction and temperature data of the cells where each wind farm is located as input, train the wind farm power output regression model, and obtain a trained wind farm power output regression model.

[0035] Wind farm power output regression models include iTransformer neural network, LSTM neural network, CNN neural network, etc.

[0036] The absolute error RMSE, mean square error NRMSE and determination coefficient R of the wind farm measured power output and the wind farm regression power output are calculated. 2 Evaluate the performance of wind farm wind power output regression model, R 2 The larger the value, the smaller the RMSE and NRMSE, indicating better performance of the model.

[0037] like Figure 1 , iTransformer is used to build a wind farm power output regression model.

[0038] iTransformer structure is as follows Figure 1 As shown in the figure, the encoder-only architecture of the transformer is adopted, including embedding, projection, and transformer blocks. The overall structure of iTransformer shares the same module arrangement as the encoder of the transformer. The original sequences of different variables are embedded as tokens independently. Self-attention is applied to the embedded variable tokens, which improves the interpretability of revealing multivariate correlations. The series representation of each token is extracted by a shared feed-forward network. Layer normalization is used to reduce the differences between variables.

[0039] Using iTransformer, wind speed, wind direction and temperature are used as input to invert the power generation output of the wind farm. Taking a wind farm in the province as an example, 75% of the data of the annual power generation output in 2021 is used as the training set, and 25% of the data is used as the test set. The feature dimensions of the encoder and decoder are both 2, the attention mechanism factor is 3, the model dimension is 128, and the feedforward network dimension is 256. Regression results and model mean absolute error RMSE, mean square error NRMSE and determination coefficient R 2 As shown in Table 2, Figure 2 shown.

[0040]

[0041] Among them, x i is the measured value of wind farm power, is the measured mean value of wind power, y i is the regression value of wind farm power, is the regression mean of wind power, and n is the number of wind farm powers.

[0042] Depend on Figure 2 It can be seen that the scattered points are closely around the diagonal line, and the model has a high degree of fit and credibility. As shown in Table 2, the evaluation indicators also verify the fitting ability and accuracy of the model. The model can better invert the historical power output data of the wind farm.

[0043] Table 2 Evaluation indicators

[0044] Evaluation indicators result Mean absolute error RMSE 0.089 Mean square error NRMSE 0.048 <![CDATA[Coefficient of determination R 2 > 0.991

[0045] 4. Establish a regression model for the power generation output of all wind farms in the province one by one, input the wind speed, wind direction and temperature of the cells where each wind farm is located into the wind farm power generation output regression model, and the output result is the historical wind power generation output data of each wind farm. After generating the historical hourly power generation output data of all wind farms for many years, accumulate the historical wind power generation output data of the province for many years.

[0046] Conventional wind power output forecasting is based on the output data of the current wind farm or the weather forecast data of the future wind farm location to predict future wind power output. Different from wind power output forecasting, this method is to obtain wind power output data under historical meteorological conditions. There are two main differences:

[0047] First, the existing wind farm power output prediction uses the local weather data of the wind farm to predict the future, while this method uses the historical weather data.

[0048] Secondly, the wind farm power output forecast uses local data of the wind farm, but there is no historical meteorological observation data of the wind farm for decades, so this method uses REA5 meteorological reanalysis data. ERA5 stores historical meteorological data in 0.25° grids (about 30 km × 30 km) of longitude and latitude. The wind farm power output model training and inversion of historical wind power output use ERA5 data obtained from the grid corresponding to the longitude and latitude of the wind farm.

[0049] Other current studies that analyze wind power output in vast areas based on ERA5 meteorological data do not consider the actual location of wind farms. Instead, they are based on the wind resource conditions of each grid and believe that wind farms should be located more in areas with better wind resource conditions to set the weight of each grid. It is generally believed that grids with better wind resource conditions have more wind farms and are given higher weights, while grids with poorer wind resource conditions are set to have no wind farms and set the weight to zero. On this basis, the historical wind power output data of the province for many years is inverted. In fact, the site selection of wind farms does not only consider wind resource conditions, but also the environmental noise during the operation of wind turbines, grid access conditions, etc. Judging from the actual location of wind farms in provincial power grids, there is a relative deviation from the wind resource conditions.

[0050] This method is based on the actual wind farm locations in the province, and establishes wind farm power output regression models for actual wind farms one by one. It can effectively avoid the inversion deviation of wind power output data caused by wind farm location deviation, and can better reveal the regular characteristics of provincial wind power output under complex meteorological scenarios.

[0051] This method can invert the provincial wind power output data of the past few decades based on the ERA5 reanalysis meteorological data of the past few decades, and can provide an empirical basis for analyzing and extracting the power output laws of provincial wind power in various complex meteorological scenarios.

[0052] Another embodiment of the present invention provides an electronic device, including:

[0053] one or more processors;

[0054] A memory stores one or more programs, and when the one or more programs are executed by one or more processors, the one or more processors implement the steps of a method for inverting wind power output based on meteorological data.

[0055] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0056] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0057] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the steps of a method for inverting wind power output based on meteorological data when the computer program is executed by a processor.

[0058] The contents explained in the above embodiments should be understood as these embodiments are only used to more clearly illustrate the present invention, and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

Claims

1. A method for inverting provincial wind power output over many years based on meteorological data, characterized in that: The following steps are involved: S1: The actual locations of each wind farm in the province correspond to the cells of the ERA5 meteorological reanalysis database, and the hourly meridional wind speed and zonal wind speed at a height of 100m corresponding to the cells of the ERA5 meteorological reanalysis data are obtained; S2: Based on the ERA5 meteorological reanalysis data, calculate the wind speed and wind direction of each cell where the wind farm is located, and the wind speed V of the i-th cell i The expression is as follows: The wind direction θ of the i-th cell i The expression is as follows: S3: Input the historical hourly wind speed, wind direction and temperature data of the cells where each wind farm is located for many years into the pre-trained wind farm power output regression model, and output the historical hourly power output data of each wind farm for many years; S4: adding the historical hourly power output data of each wind farm in the same period for many years obtained in S3 to obtain the historical hourly power output data of the province for many years; Among them, u i is the meridional wind speed at a certain moment at a height of 100m in the i-th cell, v i is the latitudinal wind speed at the i-th cell at a height of 100 m at that moment, i = 1, 2…k, k is the number of wind farms in the area.

2. The method for inverting provincial wind power output over many years based on meteorological data according to claim 1 is characterized in that: The wind farm power generation output regression model includes an iTransformer neural network, an LSTM neural network, and a CNN neural network.

3. The method for inverting provincial wind power output over many years based on meteorological data according to claim 2 is characterized in that: The training process of the wind farm power generation output regression model includes: Obtain the hourly measured wind power output of the wind farm for one year and the wind speed, wind direction, and temperature of the cell where the wind farm is located; 75% of the acquired data is used as the training set, and 25% of the data is used as the test set to train the wind farm power output regression model; The absolute error RMSE, mean square error NRMSE and determination coefficient R of the wind farm measured power output and the wind farm regression power output are calculated. 2 Evaluate the performance of wind farm wind power output regression models.

4. The method for inverting provincial wind power output over many years based on meteorological data according to claim 3 is characterized in that: The power generation output regression models of all wind farms in the province are established one by one. After generating the historical hourly power generation output data of all wind farms for many years, the historical wind power generation output data of the province for many years are accumulated.

5. An electronic device, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, which, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the steps of the method according to any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that: The computer program is stored therein, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images