CN-122026320-A - Deep sea wind farm generating capacity assessment method based on ERA5 downscaling model

CN122026320ACN 122026320 ACN122026320 ACN 122026320ACN-122026320-A

Abstract

The invention provides an ERA5 downscaling model-based deep sea wind farm generating capacity assessment method, which belongs to the field of deep sea offshore wind power and comprises the steps of obtaining original wind farm data of ERA5 with preset height from sea level in the longitude and latitude range of a target sea area in a preset time range, carrying out region cutting and coordinate alignment, preprocessing the original wind farm data of ERA5, training an ERA5 downscaling model through super-resolution generation of an anti-network deep learning model frame, carrying out super-resolution processing on the original wind farm data by utilizing the ERA5 downscaling model to generate high-resolution wind farm data, arranging a virtual wind tower position according to a deep sea wind project, correcting the virtual wind tower wind farm data by utilizing measured data, and calculating wind resources and generating capacity of the deep sea wind farm. According to the method, ERA5 low-resolution wind field data is reduced in scale through a super-resolution generation countermeasure network, the original 0.25-degree resolution is improved to 0.0625 degrees, the wind energy resource evaluation specification is met, and the current investment and time period of offshore wind power field items are reduced.

Inventors

CAO YOUHUA
ZENG QINGQUAN
LAN YUFEI
Xiong Zinong
WEI JIE
FAN JIANYONG

Assignees

中国电建集团福建省电力勘测设计院有限公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The method for evaluating the power generation capacity of the deep sea wind farm based on the ERA5 downscaling model is characterized by comprising the following steps of: acquiring original wind field data of ERA5 at a preset height from sea level in a target sea area longitude and latitude range of a preset time range, performing region cutting and coordinate alignment, and preprocessing the original wind field data of ERA 5; Selecting super-resolution generation of an countermeasure network deep learning model frame, wherein the countermeasure network deep learning model frame is configured with a residual error module, an up-sampling module and a fidelity constraint module; Deep learning model training is carried out through super-resolution generation of an countermeasure network deep learning model frame, the self-adaptive learning rate optimizer is used for restraining training and fitting, and a Dropout layer is added into the deep learning model; performing super-resolution processing on the original wind field data by using an ERA5 downscaling model, generating target sea area high-resolution wind field data, and performing fidelity constraint on the high-resolution wind field data; According to the deep sea wind power project, virtual wind measuring tower positions are distributed in the high-resolution wind field, real-time matching and data correction are carried out on real wind field data of a preset height from the coastal sea level and the virtual wind measuring tower data, and wind resources and wind field generating capacity of the deep sea wind power field are calculated by utilizing the corrected high-resolution virtual wind measuring tower wind field data.
2. The method for evaluating the power generation capacity of the deep sea wind farm based on the ERA5 downscaling model according to claim 1, wherein the deep learning model is trained by combining learning rate preheating, cosine decay strategy and dynamic adjustment strategy based on verification loss; The learning rate is preheated, wherein the learning rate of the training round number of the previous preset percentage is increased from 0 to the initial preset learning rate; the cosine attenuation strategy is that the learning rate reaches the initial learning rate and is attenuated according to a cosine function curve, and finally the learning rate is attenuated to 1% of an initial value; the dynamic adjustment strategy based on the verification loss is to multiply the learning rate by a coefficient factor=0.95 when the number of training iterations of the continuous preset number of verification losses does not decrease.
3. The method for evaluating the power generation capacity of the deep sea wind farm based on the ERA5 downscaling model according to claim 1, wherein the method is characterized in that the ERA5 downscaling model is used for performing super-resolution processing on original wind farm data to generate target sea area high-resolution wind farm data, and specifically: Inputting the original wind field data of ERA5 into a deep learning model, and amplifying the original wind field data of ERA5 to a target high-resolution size through nearest neighbor up-sampling to generate a basic high-resolution skeleton; splicing the high-resolution wind field data to generate a residual error map, and processing the residual error map by using a predefined block mask, wherein only residual errors in the block mask are reserved; and adding the residual image processed by the block mask with the basic high-resolution framework to generate high-resolution wind field data.
4. The method for evaluating the power generation capacity of the far-reaching sea wind farm based on the ERA5 scale-down model according to claim 1, wherein the step of calculating the wind resource and the power generation capacity of the far-reaching sea wind farm according to the corrected high-resolution virtual wind tower wind farm data comprises the following steps: calculating turbulence intensity according to the corrected high-resolution virtual wind measuring tower wind field data, and counting the wind speed interval and the wind speed frequency of the high-resolution virtual wind measuring tower wind field data; Performing Weibull curve fitting on wind speed frequency distribution, and solving an optimal shape parameter k and a size parameter c; Calculating the maximum gust wind speed v 50,gust in 50 years of the reproduction period of the height of the preassembled hub, and selecting the type of the fan according to v 50,gust and the turbulence intensity; And calculating the power generation amount of the deep open sea wind field according to the corrected high-resolution virtual wind measuring tower wind field data, the fan model and the fan working time range.
5. The method for estimating power generation capacity of a deep sea wind farm based on ERA5 downscaling model according to claim 4, wherein the expression for performing Weibull curve fitting on wind speed frequency distribution is The method comprises the following steps: ; Wherein v is wind speed, k is a shape parameter, and c is a size parameter.
6. The method for evaluating the power generation capacity of the offshore wind farm based on the ERA5 downscaling model according to claim 4, wherein the step of calculating the maximum gust wind speed v 50,gust of the preassembled hub in the 50-year period of height reproduction comprises the following steps: Counting a year maximum wind speed sequence of nearly 30 years at a virtual wind measuring tower, calculating a reproduction period 50 year first meeting maximum wind speed by adopting Geng Beier extremum I type and Pearson III type frequency distribution functions, carrying out height conversion according to a wind shear index, and calculating a reproduction period 50 year first meeting maximum wind speed v 50 of a preassembled fan hub height; The gust coefficient C is calculated and expressed as: ; Wherein n is a frequency parameter, T is a time interval; The wind speed v 50,gust of the maximum gust in 50 years of reproduction period is calculated and expressed as follows: ; wherein σ 50 is the standard deviation of wind speed of the annual maximum wind speed sequence of approximately 30 years.
7. The method for estimating the power generation capacity of the deep sea wind farm based on the ERA5 scale-down model according to claim 4, wherein the calculation formula of the power generation capacity E th of the wind farm is as follows: ; Wherein m is the number of generator sets in a wind field, v 1 is the cut-in wind speed of the wind generator set, v 2 is the cut-out wind speed of the wind generator set, and p j (v) is the power generated by the j-th wind generator set when the wind speed is v; and fitting the wind speed probability distribution of the j-th wind power generator set to obtain Weibull distribution.
8. The method for evaluating the power generation capacity of the deep sea wind farm based on the ERA5 downscale model according to claim 1, wherein the spatial resolution of the high-resolution wind farm data is 0.0625 degrees, and the actual distance scale is 7km.
9. The method for evaluating the power generation capacity of the deep sea wind farm based on the ERA5 downscaling model according to claim 1, wherein the high-resolution wind farm data is subjected to fidelity constraint, in particular: And restoring the high-resolution wind field data to low-resolution wind field data through downsampling, comparing whether the wind field data on the same longitude and latitude are consistent with the original wind field data, and correcting the high-resolution wind field data if the wind field data are inconsistent with the original wind field data, so that the downsampled low-resolution wind field data are consistent with the original wind field data.
10. A computer-readable storage medium storing computer-executable instructions for performing a method of estimating power generation of a deep sea wind farm based on an ERA5 downscaling model according to any of claims 1 to 9.

Description

Deep sea wind farm generating capacity assessment method based on ERA5 downscaling model Technical Field The invention belongs to the field of deep-open sea offshore wind power, and particularly relates to a deep-open sea wind farm generating capacity assessment method based on an ERA5 scale-down model. Background ERA5 is a fifth generation global climate analysis data set issued by the European middle weather forecast center (ECMWF), provides global climate time value data, has the spatial resolution of 0.25 degrees and the actual distance scale of 25-30 km, has good continuity and regularity, and is widely used for the current period wind resource and theoretical generating capacity evaluation of offshore wind power items. According to the specification of the design specification of the wind power plant (GB 51096-2015), the radius of the control range for evaluating the site wind energy resources by the measured data of each wind measuring tower is preferably 10km. Because the spatial resolution of 25-30 km of ERA5 is greater than 10km required by the specification, the ERA5 data set is directly used for wind energy resource evaluation and power generation performance prediction analysis, and a plurality of defects exist. The traditional wind resource evaluation method mainly depends on traditional wind measurement data, but as the offshore wind power project is changed from offshore to deep open sea, the problem of lack of meteorological data in the deep sea area is often faced with far from land, in addition, the observation cost of constructing an offshore wind measuring tower is high, the project early investment is large, the time period is long, and the requirements of the aging property and the economy of the engineering project of the deep open sea offshore wind power plant in the current period are difficult to meet. Therefore, how to utilize the deep learning frame model to generate effective wind resource data at the position of the virtual wind measuring tower through computer simulation and data analysis solves the problem that the meteorological data of the deep open sea offshore wind power engineering is lacking, and the problem of the spatial resolution of the original ERA5 wind field data becomes an urgent reality problem. In the prior art, the method for downscaling by using a deep learning technology is different, the current deep learning technology has achieved remarkable results in the field of image processing, but regular ERA5 wind field data is taken as image data, and the application of a generation countermeasure network and a super-resolution technology to the downscaling of the ERA5 wind field data by using a deep learning framework is still to be studied. Most of the existing researches are limited to local optimization technologies such as downscaling, wind resource evaluation or power generation amount calculation, and the like, the whole process digital technical flow from low-resolution ERA5 wind field data to high-resolution wind field reconstruction, wind resource and power generation amount evaluation and finally achievement visualization and interaction cannot be effectively established. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a deep sea wind farm generating capacity assessment method based on an ERA5 scale-down model, ERA5 low-resolution wind farm data is scaled down by a super-resolution generation countermeasure network enhanced by a deep learning model frame, a virtual wind tower is established, wind resources and generating capacity assessment results are provided for the deep sea wind farm, and current period investment and time period of offshore wind farm items are reduced. The technical scheme of the invention is as follows: in a first aspect, the invention provides a method for evaluating the power generation capacity of a deep sea wind farm based on an ERA5 scale-down model, which comprises the following steps: acquiring original wind field data of ERA5 at a preset height from sea level in a target sea area longitude and latitude range of a preset time range, performing region cutting and coordinate alignment, and preprocessing the original wind field data of ERA 5; Selecting super-resolution generation of an countermeasure network deep learning model frame, wherein the countermeasure network deep learning model frame is configured with a residual error module, an up-sampling module and a fidelity constraint module; Deep learning model training is carried out through super-resolution generation of an countermeasure network deep learning model frame, the self-adaptive learning rate optimizer is used for restraining training and fitting, and a Dropout layer is added into the deep learning model; performing super-resolution processing on the original wind field data by using an ERA5 downscaling model, generating target sea area high-resolution wind field data, and performing fidelity constraint on the hi