CN-121614543-B - Ground-based laser radar near-ground wind field modeling method and system based on data fusion

CN121614543BCN 121614543 BCN121614543 BCN 121614543BCN-121614543-B

Abstract

The application provides a ground-based laser radar near-ground wind field modeling method and a ground-based laser radar near-ground wind field modeling system based on data fusion, and relates to the technical field of wind field modeling, wherein the method comprises the steps of firstly, utilizing a ground-based laser radar to conduct various scans on a target terrain area so as to obtain an observation data set; the method comprises the steps of obtaining earth surface temperature data which are matched with three-dimensional point cloud data in a space-time mode, fusing the earth surface temperature data with the three-dimensional point cloud data to generate a coupling data set, generating initial wind field feature vectors at different moments based on the coupling data set, processing the vectors through a long-short-term memory network to generate target wind field feature vectors, inputting the target vectors into a wind field inversion model constructed based on a high-order Taylor expansion polynomial to calculate three-dimensional wind field data and turbulence feature data, and finally constructing a wind field model of a target terrain area near the ground according to the data through a self-adaptive interpolation algorithm. The application improves the fineness and accuracy of modeling of the wind field near the ground in the complex terrain.

Inventors

WANG XUN
LIU YAN
HU ZHUANG

Assignees

华信科创(北京)科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260202

Claims (8)

1. A foundation laser radar near-ground wind field modeling method based on data fusion is characterized by comprising the following steps: scanning the target terrain area by at least one foundation laser radar in a plurality of scanning modes to obtain an observation data set of the target terrain area; Acquiring earth surface temperature data which are matched with each three-dimensional point cloud data in the observation data in a space-time mode, and carrying out data fusion on the earth surface temperature data and the corresponding three-dimensional point cloud data to generate a coupling data set; Generating initial wind field feature vectors at different moments based on the coupling data set; Generating a target wind field feature vector according to the initial wind field feature vectors at different moments by combining a long-period memory network; inputting the target wind field feature vector into a wind field inversion model, and processing the target wind field feature vector through the wind field inversion model to obtain three-dimensional wind field data and turbulence feature data, wherein the wind field inversion model is constructed based on a high-order Taylor expansion polynomial; Constructing a wind field model of the target terrain area near the ground through a self-adaptive interpolation algorithm based on the three-dimensional wind field data and the turbulence characteristic data; the generating a target wind field feature vector according to the initial wind field feature vector at different moments and in combination with a long-term and short-term memory network comprises the following steps: Sorting the initial wind field feature vectors at different moments according to a time sequence to obtain an input feature set; The input feature set is input to a memory unit of the long-short-period memory network, and the memory unit calculates unit state data at the current moment according to the input feature set and history feature information stored in the memory unit; Generating a characteristic output at the current moment based on the unit state data; combining and converting the characteristic outputs at a plurality of continuous moments to generate a target wind field characteristic vector; the memory unit calculates the unit state data at the current moment according to the input feature set and the history feature information stored in the memory unit, and the method comprises the following steps: The memory unit receives an input feature vector at the current moment in the input feature set, and inputs the input feature vector and unit state data at the last moment to a first control module of the memory unit; Calculating, by the first control module, a forgetting weight value based on the input feature vector and the unit state data at the previous moment, and performing a dot product operation on the forgetting weight value and the unit state data at the previous moment to obtain first intermediate state data; The input feature vector at the current moment and the unit state data at the last moment are input to a second control module and a third calculation module in the memory unit, wherein an update weight value is calculated through the second control module, and candidate state data is calculated through the third calculation module; And performing dot multiplication operation on the update weight value and the candidate state data through a state update module of the memory unit to obtain second intermediate state data, and performing addition operation on the first intermediate state data and the second intermediate state data to obtain unit state data at the current moment.
2. The method of claim 1, wherein the inputting the target wind farm feature vector into a wind farm inversion model, and processing the target wind farm feature vector through the wind farm inversion model, results in three-dimensional wind farm data and turbulence feature data, comprises: inputting the target wind field feature vector to an input module of a wind field inversion model, analyzing the target wind field feature vector, and separating a first input parameter and a second input parameter; carrying out deep processing on the first input parameters through a residual error network module of the wind field inversion model to obtain first deep characteristic parameters; the second input parameters are subjected to weighted fusion processing through an attention mechanism module of the wind field inversion model, so that second deep characteristic parameters are obtained; calculating the first deep characteristic parameters through a first physical relation function of a first calculation module in the wind field inversion model to generate a first calculation result of each space point; Calculating the second deep characteristic parameters through a second physical relation function of a second calculation module in the wind field inversion model to generate a second calculation result of each space point; and summarizing the first calculation results of all the space points to form three-dimensional wind field data, and summarizing the second calculation results of all the space points to form turbulence characteristic data.
3. The method of claim 1, wherein the constructing a wind field model of the target terrain area near the ground by an adaptive interpolation algorithm based on the three-dimensional wind field data and the turbulence feature data comprises: Determining a space grid of the area to be modeled according to the digital elevation model of the target terrain area; taking the space coordinates of each space point in the three-dimensional wind field data as a first interpolation node, and taking the space coordinates of each space point in the turbulence characteristic data as a second interpolation node; adaptively selecting an interpolation calculation mode of each space grid based on the space density distribution of the first interpolation node and the second interpolation node; according to the interpolation calculation mode, respectively calculating a three-dimensional wind speed vector and turbulence intensity information corresponding to each space grid; And constructing a wind field model of the target terrain area near the ground based on the three-dimensional wind speed vector and the turbulence intensity information of each space grid.
4. The method of claim 1, wherein the data fusing the surface temperature data with corresponding three-dimensional point cloud data to generate a coupled dataset comprises: marking each three-dimensional point cloud data in the observation data set as a space point, and determining the space coordinate of each space point; Extracting a corresponding temperature value from the surface temperature data according to the space coordinates of each space point, and giving a temperature attribute to each space point based on the temperature value so as to form a space point set; And carrying out joint coding on the three-dimensional point cloud data and the space coordinates and the temperature attribute in the space point set to generate a coupling data set.
5. The method of claim 1, wherein generating initial wind farm feature vectors for different moments based on the coupling dataset comprises: extracting a space-time subset corresponding to each moment in the coupling data set, and respectively executing the following steps for each space-time subset: inputting the space-time subsets into a convolutional neural network, and generating first spatial distribution features of a spatial point set based on spatial coordinates of the space-time subsets and three-dimensional point cloud data through a first feature extraction layer of the convolutional neural network; through a second feature extraction layer of the convolutional neural network, correlating the first spatial distribution feature with the temperature attribute of the space-time subset, and operating the correlated data to generate a second spatial distribution feature; And performing transformation operation on the second spatial distribution characteristic through an output layer of the convolutional neural network, and outputting an initial wind field characteristic vector.
6. A foundation laser radar near-ground wind field modeling system based on data fusion is characterized by comprising: The scanning module is used for scanning the target terrain area by utilizing at least one foundation laser radar in a plurality of scanning modes to obtain an observation data set of the target terrain area; The acquisition module is used for acquiring earth surface temperature data which are matched with each three-dimensional point cloud data in the observation data in a space-time mode, and carrying out data fusion on the earth surface temperature data and the corresponding three-dimensional point cloud data to generate a coupling data set; the first generation module is used for generating initial wind field feature vectors at different moments based on the coupling data set; the second generation module is used for generating a target wind field feature vector according to the initial wind field feature vectors at different moments and combining a long-term and short-term memory network; The input module is used for inputting the target wind field feature vector into a wind field inversion model, and processing the target wind field feature vector through the wind field inversion model to obtain three-dimensional wind field data and turbulence feature data, wherein the wind field inversion model is constructed based on a high-order Taylor expansion polynomial; the construction module is used for constructing a wind field model of the target terrain area near the ground through a self-adaptive interpolation algorithm based on the three-dimensional wind field data and the turbulence characteristic data; the generating a target wind field feature vector according to the initial wind field feature vector at different moments and in combination with a long-term and short-term memory network comprises the following steps: Sorting the initial wind field feature vectors at different moments according to a time sequence to obtain an input feature set; The input feature set is input to a memory unit of the long-short-period memory network, and the memory unit calculates unit state data at the current moment according to the input feature set and history feature information stored in the memory unit; Generating a characteristic output at the current moment based on the unit state data; combining and converting the characteristic outputs at a plurality of continuous moments to generate a target wind field characteristic vector; the memory unit calculates the unit state data at the current moment according to the input feature set and the history feature information stored in the memory unit, and the method comprises the following steps: The memory unit receives an input feature vector at the current moment in the input feature set, and inputs the input feature vector and unit state data at the last moment to a first control module of the memory unit; Calculating, by the first control module, a forgetting weight value based on the input feature vector and the unit state data at the previous moment, and performing a dot product operation on the forgetting weight value and the unit state data at the previous moment to obtain first intermediate state data; The input feature vector at the current moment and the unit state data at the last moment are input to a second control module and a third calculation module in the memory unit, wherein an update weight value is calculated through the second control module, and candidate state data is calculated through the third calculation module; And performing dot multiplication operation on the update weight value and the candidate state data through a state update module of the memory unit to obtain second intermediate state data, and performing addition operation on the first intermediate state data and the second intermediate state data to obtain unit state data at the current moment.
7. An electronic device, comprising: A memory for storing a computer program; A processor for implementing the steps of the data fusion based ground lidar near-ground wind field modeling method according to any of claims 1 to 5 when executing the computer program.
8. A computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program is executed by a processor, the computer program can implement the method for modeling the ground-based lidar near-surface wind field based on data fusion according to any one of claims 1 to 5.

Description

Ground-based laser radar near-ground wind field modeling method and system based on data fusion Technical Field The application relates to the technical field of wind field modeling, in particular to a foundation laser radar near-ground wind field modeling method and system based on data fusion. Background The foundation laser radar plays an important role in near-ground wind field observation, can acquire wind measurement data with high space-time resolution, and has positive application prospects for wind energy resource evaluation and atmospheric environment monitoring. In the prior art, in order to improve the accuracy of wind field modeling, the related methods generally introduce multiple data for fusion, for example, some schemes combine laser radar wind measurement data with weather station observation or numerical weather forecast data, complement each other at the data level, and other schemes focus on processing wind field data by using a time sequence analysis method so as to capture the time sequence change characteristics of wind speed. These methods aim at building a more accurate and reliable wind field model. However, when modeling is performed on a wind field fine structure under complex terrain, the model still has room for improving the precision of the wind field key physical process, and meanwhile, the deep association features of multi-source heterogeneous data in the space-time dimension are not fully utilized. Therefore, the modeling fineness of the near-ground wind field under the complex terrain in the prior art needs to be further improved. Disclosure of Invention The application provides a foundation laser radar near-ground wind field modeling method and system based on data fusion, which are used for solving the problems of low fineness and low accuracy of the complex terrain near-ground wind field modeling in the prior art. In order to solve the technical problems, in a first aspect, the application provides a ground-based laser radar near-ground wind field modeling method based on data fusion, comprising the following steps: scanning the target terrain area by at least one foundation laser radar in a plurality of scanning modes to obtain an observation data set of the target terrain area; Acquiring earth surface temperature data which are matched with each three-dimensional point cloud data in the observation data in a space-time mode, and carrying out data fusion on the earth surface temperature data and the corresponding three-dimensional point cloud data to generate a coupling data set; Generating initial wind field feature vectors at different moments based on the coupling data set; Generating a target wind field feature vector according to the initial wind field feature vectors at different moments by combining a long-period memory network; inputting the target wind field feature vector into a wind field inversion model, and processing the target wind field feature vector through the wind field inversion model to obtain three-dimensional wind field data and turbulence feature data, wherein the wind field inversion model is constructed based on a high-order Taylor expansion polynomial; And constructing a wind field model of the target terrain area near the ground through a self-adaptive interpolation algorithm based on the three-dimensional wind field data and the turbulence characteristic data. Optionally, the generating the target wind field feature vector according to the initial wind field feature vector at different moments and in combination with the long-term and short-term memory network includes: Sorting the initial wind field feature vectors at different moments according to a time sequence to obtain an input feature set; The input feature set is input to a memory unit of the long-short-period memory network, and the memory unit calculates unit state data at the current moment according to the input feature set and history feature information stored in the memory unit; Generating a characteristic output at the current moment based on the unit state data; and combining and converting the characteristic outputs at a plurality of continuous moments to generate a target wind field characteristic vector. Optionally, the calculating, by the memory unit, the unit state data at the current moment according to the input feature set and the history feature information stored in the memory unit includes: The memory unit receives an input feature vector at the current moment in the input feature set, and inputs the input feature vector and unit state data at the last moment to a first control module of the memory unit; Calculating, by the first control module, a forgetting weight value based on the input feature vector and the unit state data at the previous moment, and performing a dot product operation on the forgetting weight value and the unit state data at the previous moment to obtain first intermediate state data; The input feature vector at the current moment and th