CN-122020073-A - Wind field prediction method and system based on unmanned aerial vehicle cluster cooperation

CN122020073ACN 122020073 ACN122020073 ACN 122020073ACN-122020073-A

Abstract

The invention discloses a wind field prediction method and a wind field prediction system based on unmanned aerial vehicle cluster cooperation, and relates to the technical field of aircrafts, wherein the method comprises the following steps of dividing a target area into a plurality of three-dimensional grid nodes and corresponding perception domains, and mapping a plurality of key features corresponding to each unmanned aerial vehicle to the corresponding three-dimensional grid nodes according to preset rules; the method comprises the steps of inputting network nodes, node characteristics and edge weights into a graph neural network to construct a real-time global wind field characteristic graph of a target area, and inputting the global wind field characteristic graph in a current set time period into a pre-trained wind field prediction model to obtain a three-dimensional wind field predicted value of the target area in a future set time period. The global wind field feature map provided by the invention has the space dimension corresponding to the target area and the channel dimension corresponding to the wind field feature, so that a collaborative sensing network covering a low-altitude wide area is formed, and the wind field prediction result precision of the target area is improved.

Inventors

XU XIAOPING
HE YINGLONG
ZHOU ZHOU
ZHANG ZIYANG

Assignees

西北工业大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (9)

1. The wind field prediction method based on unmanned aerial vehicle cluster cooperation is characterized by comprising the following steps of: dividing a target area into a plurality of three-dimensional grid nodes and corresponding perception domains, deploying unmanned aerial vehicle clusters in the target area, and enabling unmanned aerial vehicles to fly in the plurality of perception domains according to a preset track; The method comprises the steps of acquiring flight parameters of each unmanned aerial vehicle in real time, inverting three-dimensional wind speed vectors of a current area where the unmanned aerial vehicle is located based on the flight parameters of the unmanned aerial vehicle in real time, extracting features of each three-dimensional wind speed vector to obtain a plurality of key features for representing wind field uncertainty; Taking a plurality of three-dimensional grid nodes as network nodes, taking a plurality of key features of each three-dimensional grid node as node features, acquiring edge weights among the network nodes through a Gaussian kernel function based on three-dimensional space distances among different three-dimensional grid nodes, inputting the network nodes, the node features and the edge weights into a graph neural network, and constructing a real-time global wind field feature graph of a target area; and inputting the global wind field characteristic diagram in the current set time period into a pre-trained wind field prediction model to obtain a three-dimensional wind field predicted value of the target area in the future set time period.
2. The wind field prediction method based on unmanned aerial vehicle cluster coordination according to claim 1, wherein the unmanned aerial vehicle flight parameters comprise scalar airspeed, flight attitude angle and ground speed vectors, and the three-dimensional wind speed vectors are inverted in real time based on the unmanned aerial vehicle flight parameters, and specifically comprises the following steps: constructing a rotation matrix from a body coordinate system to an inertial coordinate system based on the flight attitude angle; inputting scalar airspeed of each position into a rotation matrix to obtain an airspeed vector under an inertial coordinate system; and (3) performing difference on the ground speed vector and the airspeed vector to obtain a real-time three-dimensional wind speed vector at each position.
3. The wind field prediction method based on unmanned aerial vehicle cluster coordination as claimed in claim 2, wherein the airspeed vector under the inertial coordinate system is specifically as follows: ; In the formula, Is the airspeed vector in the inertial coordinate system, In order to rotate the matrix is rotated, B refers to an engine body coordinate system, i refers to an inertial coordinate system; the three-dimensional wind speed vector is specifically shown as follows: ; In the formula, 、 And Is the east, north and vertical components of airspeed in the inertial frame, 、 And Is the east, north and vertical components of the three-dimensional wind speed vector in the inertial coordinate system, 、 And Is the east, north and vertical components of ground speed.
4. The wind field prediction method based on unmanned aerial vehicle cluster coordination according to claim 2 is characterized in that the unmanned aerial vehicle flight parameters further comprise longitude, latitude and altitude, and the method specifically comprises the following steps of: Projecting longitude and latitude to an east coordinate and a north coordinate under an inertial coordinate system, and taking the altitude as a vertical coordinate under the inertial coordinate system; Acquiring horizontal wind speed gradients through the east and north components of the three-dimensional wind speed vector under the inertial coordinate system and the east and north coordinates; acquiring vertical wind shear through vertical components and vertical coordinates of a three-dimensional wind speed vector under an inertial coordinate system; Acquiring turbulence intensity through the east component of the three-dimensional wind speed vector under the inertial coordinate system; and obtaining the wind speed time change rate through a model of the three-dimensional wind speed vector.
5. The wind field prediction method based on unmanned plane cluster cooperation of claim 1, wherein the wind field prediction model is a CNN-LSTM model and comprises a CNN layer and an LSTM layer, the global wind field feature map in the current set time period is input into a pre-trained wind field prediction model to obtain a three-dimensional wind field prediction value of each position future set time period, and the method comprises the following steps: Acquiring an atmospheric turbulence theoretical model and historical wind field data of a target area; In the characteristic extraction process, taking physical constraint parameters of an atmospheric turbulence theoretical model as regular terms, and embedding a weight updating process of a convolution kernel; And in the time sequence capturing process, the forgetting door weight of the LSTM unit is adjusted through historical wind field data.
6. The wind field prediction method based on unmanned aerial vehicle cluster cooperation as claimed in claim 1, wherein the preset rule specifically comprises: The unmanned aerial vehicle acquires the position coordinates of the unmanned aerial vehicle in real time, acquires the actual distances between the unmanned aerial vehicle and all three-dimensional grid node coordinates, and matches the three-dimensional grid node which is closest to the actual distances and is in the perception domain of the unmanned aerial vehicle based on the actual distances, wherein a plurality of key features of the unmanned aerial vehicle are preferentially attributed to the three-dimensional grid node; If the position of the unmanned aerial vehicle is in the perception domain overlapping area of the three-dimensional grid nodes, three-dimensional space distance between the unmanned aerial vehicle and each overlapping three-dimensional grid node is obtained, data weight is distributed according to an inverse distance weighting method, and a plurality of key features are distributed to each overlapping three-dimensional grid node according to the weight.
7. The unmanned aerial vehicle cluster cooperation-based wind field prediction method of claim 1, wherein wind field prediction model parameters are updated by adaptive kalman filtering.
8. The unmanned aerial vehicle cluster collaboration-based wind field prediction method of claim 1, wherein unmanned aerial vehicle flight parameters at each location are time synchronized by kalman filtering.
9. Wind field prediction system based on unmanned aerial vehicle cluster cooperation, characterized by comprising: the deployment module is used for dividing the target area into a plurality of three-dimensional grid nodes and corresponding perception domains, deploying unmanned aerial vehicle clusters in the target area, and enabling the unmanned aerial vehicle to fly in the plurality of perception domains according to a preset track; The system comprises an extraction module, a feature extraction module, a three-dimensional grid node, a wind field uncertainty characterization module, a three-dimensional grid node, a wind field uncertainty analysis module and a three-dimensional grid node, wherein the extraction module is used for acquiring flight parameters of each unmanned aerial vehicle in real time, inverting a three-dimensional wind speed vector of a current area where the unmanned aerial vehicle is located based on the flight parameters of the unmanned aerial vehicle in real time, extracting features of each three-dimensional wind speed vector, and obtaining a plurality of key features for characterizing the wind field uncertainty; the construction module is used for taking a plurality of three-dimensional grid nodes as network nodes, taking a plurality of key features of each three-dimensional grid node as node features, acquiring edge weights among the network nodes through a Gaussian kernel function based on three-dimensional space distances among different three-dimensional grid nodes, inputting the network nodes, the node features and the edge weights into a graph neural network, and constructing a real-time global wind field feature graph of a target area; the prediction module is used for inputting the global wind field characteristic diagram in the current set time period into a pre-trained wind field prediction model to obtain a three-dimensional wind field predicted value of the future set time period of the target area.

Description

Wind field prediction method and system based on unmanned aerial vehicle cluster cooperation Technical Field The invention relates to the technical field of aircrafts, in particular to a wind field prediction method and system based on unmanned aerial vehicle cluster cooperation. Background The low-altitude economy is an emerging economic form with wide radiation range and industrial chain length, and has become an important direction of high-quality development of economy in China. The low-altitude economy covers a plurality of fields such as unmanned aerial vehicle logistics, emergency rescue, agricultural plant protection and urban traffic, the core of the low-altitude economy depends on safe and efficient flight activities, the application scenes provide extremely high requirements on the flight safety and efficiency of the unmanned aerial vehicle, and the accurate perception and prediction of the flight environment are the basis for realizing the situation. Meanwhile, the accurate wind field prediction can support dynamic optimization of the flight path, and the unmanned aerial vehicle can actively select the downwind route according to the prediction information and avoid a strong upwind area, so that the energy consumption is obviously reduced, the endurance mileage is prolonged, the task execution time is shortened, and the low-altitude economy is promoted to be in a mature and intelligent stage from primary application. However, the low-altitude airspace has complex meteorological conditions and dynamic obstacles (such as trees, buildings, other aircrafts and the like), so that the low-altitude airspace has the characteristics of strong burst, complex space-time variation, multi-factor influence, unsteadiness and nonlinearity, local and regional differences, uncertainty under complex meteorological conditions and the like. These characteristics make the perception and prediction of low-altitude wind farms a great challenge, requiring high-precision, high-spatial-temporal resolution perception techniques to cope with. The existing wind field prediction method collects wind field parameters of a target area through fixed sensors (such as weather stations and wind profile radars) or a single unmanned aerial vehicle, only local small-range wind field perception can be realized, comprehensive and accurate perception of a wide-area dynamic wind field cannot be covered, and the wind field prediction result of the target area is low in precision. Disclosure of Invention Based on the defects in the prior art, the invention provides a wind field prediction method and a wind field prediction system based on unmanned aerial vehicle cluster cooperation, which solve the existing problems. The invention adopts the following technical scheme: In a first aspect, the invention provides a wind field prediction method based on unmanned aerial vehicle cluster cooperation, which comprises the following steps: dividing a target area into a plurality of three-dimensional grid nodes and corresponding perception domains, deploying unmanned aerial vehicle clusters in the target area, and enabling unmanned aerial vehicles to fly in the plurality of perception domains according to a preset track; The method comprises the steps of acquiring flight parameters of each unmanned aerial vehicle in real time, inverting three-dimensional wind speed vectors of a current area where the unmanned aerial vehicle is located based on the flight parameters of the unmanned aerial vehicle in real time, extracting features of each three-dimensional wind speed vector to obtain a plurality of key features for representing wind field uncertainty; Taking a plurality of three-dimensional grid nodes as network nodes, taking a plurality of key features of each three-dimensional grid node as node features, acquiring edge weights among the network nodes through a Gaussian kernel function based on three-dimensional space distances among different three-dimensional grid nodes, inputting the network nodes, the node features and the edge weights into a graph neural network, and constructing a real-time global wind field feature graph of a target area; and inputting the global wind field characteristic diagram in the current set time period into a pre-trained wind field prediction model to obtain a three-dimensional wind field predicted value of the target area in the future set time period. Preferably, the unmanned aerial vehicle flight parameters comprise scalar airspeed, flight attitude angle and ground speed vectors, and the three-dimensional wind speed vector is inverted in real time based on the unmanned aerial vehicle flight parameters, and specifically comprises the following steps: constructing a rotation matrix from a body coordinate system to an inertial coordinate system based on the flight attitude angle; inputting scalar airspeed of each position into a rotation matrix to obtain an airspeed vector under an inertial coordinate system; and (3) per