CN-121997028-A - Bending moment load prediction method and device for offshore wind power support structure

CN121997028ACN 121997028 ACN121997028 ACN 121997028ACN-121997028-A

Abstract

The invention relates to the technical field of wind power generation, in particular to a bending moment load prediction method and device of an offshore wind power support structure, comprising the steps of collecting running state parameters and actual acceleration at a target position in real time; the method comprises the steps of inputting the input data into a trained bending moment load prediction model to obtain bending moment load prediction data of a tower base section, wherein the model is trained by carrying out local feature extraction and fusion of time windows of different lengths on the input data to obtain first enhancement features, splicing and arranging the input data according to the same time stamp by simulated acceleration and simulated operation parameters, combining a first position feature vector with the first enhancement features to obtain second time sequence features, obtaining a predicted bending moment time sequence signal based on the second time sequence features through variable attention calculation and linear mapping, constructing a loss function model, and training to obtain the trained bending moment load prediction model. According to the technical scheme, the prediction precision of the bending moment load of the section of the tower base can be improved.

Inventors

LI GANG
XU MINGQIANG
LIU BO
Ma chunke
LIU ZHENHAI
WANG SHUQING
TANG DONGLIANG

Assignees

中国电力工程顾问集团有限公司
中国海洋大学

Dates

Publication Date: 20260508
Application Date: 20260209

Claims (10)

1. A method for predicting bending moment load of an offshore wind power support structure is characterized by comprising the following steps: acquiring running state parameters of the wind turbine generator and actual acceleration of a target position on a wind power supporting structure in real time; Inputting the running state parameters and the actual acceleration into a trained bending moment load prediction model to obtain output bending moment load prediction data of the cross section of the tower base of the wind power support structure; The bending moment load prediction model is trained by the following modes: based on the plurality of simulated wind condition data and the plurality of simulated wave data, obtaining simulated operation parameters of the wind turbine, simulated acceleration of a target position and simulated bending moment load of a tower base section; Extracting local features of time windows with different lengths from input data in a time dimension, fusing the extracted multiple parallel local features to obtain a first enhancement feature, and splicing the input data according to the same time stamp by the simulated acceleration and the simulated operation parameter and then arranging the input data according to the sequence of the time stamp; Combining a first position feature vector with the first enhanced feature to obtain a second time sequence feature, wherein the first position feature vector is generated based on the dynamic response characteristic of the offshore wind power support structure; Based on the second time sequence characteristic, obtaining a predicted bending moment time sequence signal of the section of the tower base through variable attention calculation and linear mapping; And training the model based on a loss function constructed based on the difference between the predicted bending moment time sequence signal and the simulated bending moment load to obtain a trained bending moment load prediction model.
2. The method of claim 1, wherein combining the first location feature vector with the first enhancement feature results in a second timing feature, comprising: Performing a leachable position code on the first enhancement feature to generate the first position feature vector, wherein the leachable position code is iteratively optimized with the aim of minimizing the loss function in the training process of the bending moment load prediction model, and the leachable position code is used for representing the position feature vector of the dynamic response characteristic of the wind power support structure in different operation stages, wherein the different operation stages comprise a preset starting stage, a rated operation stage, a turbulence working condition stage and a stopping stage; And combining the first position feature vector with the first enhanced feature to obtain a second time sequence feature.
3. The method of claim 1, wherein the obtaining the predicted bending moment time sequence signal of the tower base section based on the second time sequence feature through variable attention calculation and linear mapping comprises: Mapping the second timing characteristic into a query vector Q, a key vector K, and a value vector V; Generating an attention weight matrix between the query vector Q and the key vector K based on the correlation between the query vector Q and the key vector K; Carrying out weighted fusion on the value vector V based on the attention weight matrix to obtain deep features; and carrying out linear mapping on the deep layer characteristics on the full-connection layer to obtain a predicted bending moment time sequence signal of the section of the tower base.
4. The method according to claim 1, wherein the performing local feature extraction of time windows of different lengths on the input data in the time dimension, and fusing the extracted multiple parallel local features to obtain the first enhancement feature, includes: extracting local features of the input data by using a plurality of parallel time windows with different lengths; carrying out nonlinear activation and random inactivation regularization treatment on the local features extracted from each time window to obtain a plurality of regularized local features; and performing splicing treatment on the regularized local features to obtain a first enhancement feature.
5. The method of claim 1, wherein the bending moment load prediction data comprises bending moment loads in an x-direction and bending moment loads in a y-direction of a tower base section of the wind power support structure; the simulated bending moment load comprises bending moment load of the wind power support structure tower barrel base section in the x direction and bending moment load in the y direction.
6. The method of claim 5, wherein the operating state parameters include x-direction wind speed, y-direction wind speed, yaw angle, rotational speed, and pitch angle.
7. The method of claim 1, wherein the obtaining the simulated operating parameters of the wind turbine, the simulated acceleration of the target location, and the simulated bending moment load of the tower base section based on the plurality of simulated wind condition data and the plurality of simulated wave data comprises: Generating simulated wind condition data by using IEC KAIMAL spectrum models, and generating corresponding simulated wave data by using JONSWAP spectrum models; Based on the simulated wind condition data and the simulated wave data, a OpenFAST model is utilized to obtain simulated operation parameters of the wind turbine generator, simulated acceleration of a target position and simulated bending moment load of a tower base section.
8. The utility model provides a wind-powered electricity generation bearing structure's at sea moment load prediction device which characterized in that includes: the data acquisition module is used for acquiring the running state parameters of the wind turbine generator and the actual acceleration of a target position on the wind power support structure in real time; The bending moment prediction module is connected with the data acquisition module, and inputs the running state parameters and the actual acceleration into a trained bending moment load prediction model to obtain output bending moment load prediction data of the cross section of the tower base of the wind power support structure; The bending moment load prediction model is trained by the following modes: based on the plurality of simulated wind condition data and the plurality of simulated wave data, obtaining simulated operation parameters of the wind turbine, simulated acceleration of a target position and simulated bending moment load of a tower base section; Extracting local features of time windows with different lengths from input data in a time dimension, fusing the extracted multiple parallel local features to obtain a first enhancement feature, and splicing the input data according to the same time stamp by the simulated acceleration and the simulated operation parameter and then arranging the input data according to the sequence of the time stamp; Combining a first position feature vector with the first enhanced feature to obtain a second time sequence feature, wherein the first position feature vector is generated based on the dynamic response characteristic of the offshore wind power support structure; Based on the second time sequence characteristic, obtaining a predicted bending moment time sequence signal of the section of the tower base through variable attention calculation and linear mapping; And training the model based on a loss function constructed based on the difference between the predicted bending moment time sequence signal and the simulated bending moment load to obtain a trained bending moment load prediction model.
9. An electronic device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the method of any of claims 1-7 when the computer program is executed.
10. A computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of any of claims 1-7.

Description

Bending moment load prediction method and device for offshore wind power support structure Technical Field The invention relates to the technical field of wind power generation, in particular to a bending moment load prediction method and device for an offshore wind power support structure. Background The offshore wind power supporting structure is a core force transmission component for connecting an upper fan with a seabed foundation, and is required to bear complex alternating load under the coupling action of multiple physical fields such as wind, waves, currents and the like in the long-term service process. The section of the connecting part of the wind power tower barrel and the wind power foundation is the area with the most obvious stress concentration phenomenon, and is also a key link with weak structural safety, and the bending moment at the section is a core mechanical parameter reflecting the whole stress and power response of the offshore wind power supporting structure, so that the monitoring of the bending moment at the section for a long period and high precision is important. In the prior art, direct measurement is often used to obtain the moment load of the connection of the tower to the foundation. According to the method, physical sensors such as strain gauges are arranged at key positions such as a tower foundation or an underwater mud surface to acquire strain signals, and the strain signals are converted into bending moment loads after analysis and calculation. However, due to the characteristics of high humidity, high salt fog, strong corrosion and the like of the marine environment, the corrosion failure of the sensor can be caused, the drift of the measurement signal occurs, and the accuracy of the measurement data is reduced. In addition, the installation and maintenance of the underwater measuring points need professional offshore operation equipment and personnel, the operation difficulty is high, the cost is high, and long-term stable continuous monitoring is difficult to realize. Therefore, those skilled in the art are required to develop a new technical solution to solve the above problems. Disclosure of Invention The invention provides a bending moment load prediction method and device for an offshore wind power support structure, which can improve the prediction precision of the bending moment load of the section of a tower base. In a first aspect, an embodiment of the present invention provides a method for predicting bending moment load of an offshore wind power support structure, including: acquiring running state parameters of the wind turbine generator and actual acceleration of a target position on a wind power supporting structure in real time; Inputting the running state parameters and the actual acceleration into a trained bending moment load prediction model to obtain output bending moment load prediction data of the cross section of the tower base of the wind power support structure; The bending moment load prediction model is trained by the following modes: based on the plurality of simulated wind condition data and the plurality of simulated wave data, obtaining simulated operation parameters of the wind turbine, simulated acceleration of a target position and simulated bending moment load of a tower base section; Extracting local features of time windows with different lengths from input data in a time dimension, fusing the extracted multiple parallel local features to obtain a first enhancement feature, and splicing the input data according to the same time stamp by the simulated acceleration and the simulated operation parameter and then arranging the input data according to the sequence of the time stamp; Combining a first position feature vector with the first enhanced feature to obtain a second time sequence feature, wherein the first position feature vector is generated based on the dynamic response characteristic of the offshore wind power support structure; Based on the second time sequence characteristic, obtaining a predicted bending moment time sequence signal of the section of the tower base through variable attention calculation and linear mapping; And training the model based on a loss function constructed based on the difference between the predicted bending moment time sequence signal and the simulated bending moment load to obtain a trained bending moment load prediction model. In a second aspect, an embodiment of the present invention provides a bending moment load prediction apparatus for an offshore wind power support structure, including: the data acquisition module is used for acquiring the running state parameters of the wind turbine generator and the actual acceleration of a target position on the wind power support structure in real time; The bending moment prediction module is connected with the data acquisition module, and inputs the running state parameters and the actual acceleration into a trained bending moment load prediction