CN-122026331-A - Wind farm short-term power load prediction method and system based on digital twin

CN122026331ACN 122026331 ACN122026331 ACN 122026331ACN-122026331-A

Abstract

The invention relates to the technical field of wind farm short-term power load prediction, and discloses a wind farm short-term power load prediction method and system based on digital twinning, wherein the method comprises the steps of firstly obtaining weather forecast values, dividing the weather forecast values into layered data according to atmospheric stability after correction, mapping to the iterative evolution of the three-dimensional grid to generate an updated three-dimensional wind field, constructing a space discrete wind field by combining the updated three-dimensional wind field, wind field data and fan positions, correcting to obtain a corrected wind field, determining the wake flow influence range of the corrected wind field, and adjusting through adjusting points to obtain a final wind field. Screening out a normal fan set with wind speed exceeding a preset wind speed threshold value, fitting a change curve of wind speed and power by using a polynomial, generating a power set by a wind farm fan according to the change curve, inputting an LSTM model to obtain a predicted power set, correcting and smoothing the predicted power set to form a final power set. The method can accurately describe the wind farm and accurately reflect the power change of the wind farm fan.

Inventors

Cheng gun
LV SEN
ZOU YONG

Assignees

沈阳科技学院

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (8)

1. The method for predicting the short-term power load of the wind farm based on digital twinning is characterized by comprising the following steps: acquiring meteorological data, carrying out layering treatment on the meteorological data to obtain layering data, mapping the layering data to a three-dimensional grid, carrying out iterative evolution, and generating an updated three-dimensional wind field; Constructing a space discrete wind field according to the updated three-dimensional wind field, the acquired wind field data and the acquired fan position information, positioning a fan action area in the space discrete wind field, calculating a momentum source item of the fan action area, and correcting the space discrete wind field according to the momentum source item to obtain a corrected wind field; Determining a wake flow influence range of the corrected wind field, calculating the speed loss rate of each grid point in the wake flow influence range, marking the grid point with the speed loss rate larger than a preset loss rate threshold as an adjusting point, and adjusting the corrected wind field according to the adjusting point to obtain a final wind field; Finding out all fans with the wind speeds greater than a preset wind speed threshold value in the final wind field, integrating the fans into a normal fan set, and fitting the change rule of the historical wind speed and the power of the normal fan set by using a polynomial to obtain a change curve; Acquiring the wind speed of a wind power plant fan at each moment, inputting the change curve to obtain a power set, inputting the power set into a preset long-short-period memory neural network model for prediction, and obtaining a predicted power set; and acquiring influence fans influenced by wake flow in the predicted power set, simultaneously acquiring all wake flow source fans corresponding to the influence fans, calculating comprehensive loss, correcting and smoothing the predicted power set according to the comprehensive loss, and obtaining a final power set.
2. The method for predicting short-term power load of a wind farm based on digital twinning according to claim 1, wherein the steps of obtaining meteorological data, layering the meteorological data to obtain layered data, mapping the layered data to a three-dimensional grid and performing iterative evolution to generate an updated three-dimensional wind farm comprise: Acquiring a weather forecast value, and simultaneously, monitoring weather data on the ground to obtain an actual measurement value, and correcting the weather forecast value according to the actual measurement value to obtain a corrected weather value; calculating Mo Ningao the length of the hough of the corrected meteorological value, and evaluating and layering the atmospheric stability of the corrected meteorological value according to the Mo Ningao length of the hough to obtain layering data; Calculating to obtain vertical wind profiles of all positions according to the layered data, and mapping the vertical wind profiles to a three-dimensional grid to form an initial three-dimensional wind field; and iteratively evolving the initial three-dimensional wind field through a preset numerical solver to obtain an updated three-dimensional wind field.
3. The method for predicting short-term power load of a wind farm based on digital twinning according to claim 1, wherein the constructing a spatial discrete wind farm according to the updated three-dimensional wind farm and the acquired wind farm data and wind farm position information, locating a wind farm action region in the spatial discrete wind farm, calculating a momentum source item of the wind farm action region, and correcting the spatial discrete wind farm according to the momentum source item to obtain a corrected wind farm comprises: Acquiring wind power plant data, fan position information and fan blade geometric data; Constructing a space discrete wind field according to the updated three-dimensional wind field, the wind field data and the fan position information; Calculating a blade sweep area according to the fan position information and the fan blade geometric data, and mapping the blade sweep area to the space discrete wind field to obtain a fan action grid; and calculating a momentum source item of the fan action grid through a preset actuation disc model, and inputting the space discrete wind field and the momentum source item into a preset numerical solver to obtain a corrected wind field.
4. The method for predicting short-term power load of a wind farm based on digital twinning according to claim 1, wherein determining a wake influence range of the corrected wind farm, calculating a speed deficit rate of each grid point in the wake influence range, marking the grid point with the speed deficit rate greater than a preset deficit rate threshold as an adjustment point, and adjusting the corrected wind farm according to the adjustment point to obtain a final wind farm, comprises: determining a dominant incoming flow wind direction according to the corrected wind field, and determining a wake flow influence range according to the dominant incoming flow wind direction; Calculating the speed loss rate of each grid point in the wake flow influence range, and marking the grid point as an adjusting point if the speed loss rate is larger than a preset loss rate threshold value to obtain an adjusting point list; and finding a regulating point in the regulating point list in the corrected wind field, regulating the wind speed of the regulating point, and updating the wind speed value to obtain a final wind field.
5. The method for predicting short-term power load of a wind farm based on digital twinning according to claim 1, wherein the steps of finding out fans with wind speeds greater than a preset wind speed threshold in the final wind farm and integrating the fans into a normal fan set, fitting a change rule of historical wind speeds and power of the normal fan set by using a polynomial to obtain a change curve include: Searching a grid point corresponding to the nearest grid point of the fan position information as a fan wind speed aiming at each fan in the final wind field to obtain a wind speed sequence; if the wind speed of the fans in the wind speed sequence is greater than a preset wind speed threshold value, marking the fans as normal fans to obtain a normal fan set; And acquiring the historical wind speed and power of each fan in the normal fan set, and fitting a wind speed and power change rule by using a polynomial to obtain a change curve.
6. The method for predicting the short-term power load of a wind farm based on digital twin according to claim 1, wherein the steps of obtaining the wind speed of a wind farm fan at each moment and inputting the change curve to obtain a power set, inputting the power set into a preset long-short-term memory neural network model to predict, and obtaining a predicted power set comprise the following steps: acquiring wind field data of a wind power field fan in a preset time period, and extracting wind speeds at all times in the preset time period; inputting the wind speed at each moment into the change curve to obtain power at each moment, and integrating a time point and the power at each moment into a power set; and inputting the power set into a preset long-short-term memory neural network model for prediction to obtain a predicted power set.
7. The method for predicting the short-term power load of a wind farm based on digital twinning according to claim 1, wherein the obtaining the influence fans influenced by wake flows in the predicted power set, simultaneously obtaining all wake flow source fans corresponding to the influence fans and calculating a comprehensive deficit, correcting and smoothing the predicted power set according to the comprehensive deficit to obtain a final power set includes: acquiring a wind direction angle at each moment in the predicted power set, and calculating a wake flow influence range of a wind power plant fan according to the wind direction angle to obtain an influence fan; Acquiring a wake source fan which generates wake effect on the influence fan, calculating comprehensive loss caused by the wake source fan on the influence fan, and correcting the predicted power set according to the comprehensive loss to obtain a corrected power set; and carrying out smooth filtering on the corrected power set to obtain a final power set.
8. A system using the digital twin based wind farm short term power load prediction method according to any of claims 1 to 7, comprising: the initial wind field generation module is used for acquiring meteorological data, carrying out layering treatment on the meteorological data to obtain layering data, mapping the layering data to a three-dimensional grid, carrying out iterative evolution, and generating an updated three-dimensional wind field; the wind field correction module is used for constructing a space discrete wind field according to the updated three-dimensional wind field, the acquired wind field data and the acquired fan position information, positioning a fan action area in the space discrete wind field, calculating a momentum source item of the fan action area, and correcting the space discrete wind field according to the momentum source item to obtain a corrected wind field; The final wind field generation module is used for determining a wake flow influence range of the corrected wind field, calculating the speed loss rate of each grid point in the wake flow influence range, marking the grid point with the speed loss rate larger than a preset loss rate threshold as an adjustment point, and adjusting the corrected wind field according to the adjustment point to obtain a final wind field; the regular fitting module is used for finding out fans with the wind speeds of all fans being larger than a preset wind speed threshold value in the final wind field, integrating the fans into a normal fan set, and using polynomials to fit the change rules of the historical wind speeds and the power of the normal fan set to obtain a change curve; The power prediction module is used for acquiring the wind speed of the wind power plant fan at each moment and inputting the change curve to obtain a power set, and inputting the power set into a preset long-period and short-period memory neural network model to predict to obtain a predicted power set; the final power generation module is used for acquiring the influence fans influenced by the wake flow in the predicted power set, simultaneously acquiring all wake flow source fans corresponding to the influence fans, calculating comprehensive loss, correcting and smoothing the predicted power set according to the comprehensive loss, and obtaining a final power set.

Description

Wind farm short-term power load prediction method and system based on digital twin Technical Field The invention relates to the technical field of wind farm short-term power load prediction, in particular to a wind farm short-term power load prediction method and system based on digital twinning. Background Currently, wind power generation is an important component of clean energy and occupies an increasingly critical position in global energy conversion. The wind farm power generation is greatly influenced by natural wind resources, has strong randomness and volatility, and ensures that short-term power load prediction can reach higher prediction precision by relying on industrial data analysis technology so as to support power grid dispatching and energy optimization configuration. Under the prior art, most prediction methods often rely on historical power generation data and basic meteorological elements for statistical modeling, however, in a wind power plant with actual complex terrain and multi-wind turbine classification arrangement, the methods are difficult to accurately describe the actual distribution rule of wind speed in space in the field. Meanwhile, when the fan operates, an obvious wake flow effect is generated, so that the wind speed of the downstream fan is obviously reduced. The existing prediction means often neglect or simplify the mutual influence relation among fans, so that the real change trend of the fan power in the wind power plant in different moments is difficult to capture. Therefore, the prior art has the defects that the wind farm can not be accurately described and the change of the fan power of the wind farm can not be accurately reflected. Disclosure of Invention The invention provides a wind farm short-term power load prediction method and system based on digital twinning, which can solve the problems that wind farms cannot be accurately represented and wind farm fan power changes can be accurately reflected in the prior art. In order to solve the technical problems, the invention provides a wind farm short-term power load prediction method based on digital twinning, which comprises the following steps: acquiring meteorological data, carrying out layering treatment on the meteorological data to obtain layering data, mapping the layering data to a three-dimensional grid, carrying out iterative evolution, and generating an updated three-dimensional wind field; Constructing a space discrete wind field according to the updated three-dimensional wind field, the acquired wind field data and the acquired fan position information, positioning a fan action area in the space discrete wind field, calculating a momentum source item of the fan action area, and correcting the space discrete wind field according to the momentum source item to obtain a corrected wind field; Determining a wake flow influence range of the corrected wind field, calculating the speed loss rate of each grid point in the wake flow influence range, marking the grid point with the speed loss rate larger than a preset loss rate threshold as an adjusting point, and adjusting the corrected wind field according to the adjusting point to obtain a final wind field; Finding out all fans with the wind speeds greater than a preset wind speed threshold value in the final wind field, integrating the fans into a normal fan set, and fitting the change rule of the historical wind speed and the power of the normal fan set by using a polynomial to obtain a change curve; Acquiring the wind speed of a wind power plant fan at each moment, inputting the change curve to obtain a power set, inputting the power set into a preset long-short-period memory neural network model for prediction, and obtaining a predicted power set; and acquiring influence fans influenced by wake flow in the predicted power set, simultaneously acquiring all wake flow source fans corresponding to the influence fans, calculating comprehensive loss, correcting and smoothing the predicted power set according to the comprehensive loss, and obtaining a final power set. In an optional implementation manner, the acquiring meteorological data, performing layering processing on the meteorological data to obtain layering data, mapping the layering data to a three-dimensional grid, performing iterative evolution, and generating an updated three-dimensional wind field, and includes: Acquiring a weather forecast value, and simultaneously, monitoring weather data on the ground to obtain an actual measurement value, and correcting the weather forecast value according to the actual measurement value to obtain a corrected weather value; calculating Mo Ningao the length of the hough of the corrected meteorological value, and evaluating and layering the atmospheric stability of the corrected meteorological value according to the Mo Ningao length of the hough to obtain layering data; Calculating to obtain vertical wind profiles of all positions according to the layered data, an