Dimensionality reduction and reconstruction method of spatial multi-dimensional wind power data based on rbf kernel function

Abstract

The invention discloses a method for dimensionality reduction of spatial multi-dimensional wind power data based on RBF kernel function, comprising: collecting multiple continuous hour-level power data samples of multiple wind farms; calculating the kernel matrix; calculating the eigenvalues ​​and eigenvectors of the kernel matrix; The values ​​are sorted in descending order, and the first r eigenvalues ​​and corresponding eigenvectors in the eigenvalue sequence are taken; the dimensionality reduction result is calculated using the kernel matrix and the selected eigenvalues ​​and eigenvectors. The invention also discloses a corresponding data reconstruction method, including: taking the minimum error between the original data dimension reduction result and the constructed homogeneous data dimension reduction result as the optimization goal, searching for the optimal kernel parameter; The result is refactored. The method of the invention makes the data less and the features more independent, which is conducive to the realization of grid-connected operation of large-scale wind power data; the dimensionality reduction data of the invention is superior to the existing dimensionality reduction method in terms of reliability and continuity; The error between the reconstructed data obtained by the inventive method and the original data is small.

Description

technical field [0001] The invention belongs to the field of renewable energy power generation and comprehensive consumption, and in particular relates to a method for dimensionality reduction and reconstruction of spatial multi-dimensional wind power data based on RBF kernel functions. Background technique [0002] As a clean, low-cost and technologically mature renewable energy, wind energy has become the largest form of new energy generation currently developed. However, due to the strong randomness and intermittent nature of wind energy, a large number of wind power connected to the grid will inevitably have a huge impact on the security and stable operation of the grid. At present, many scholars have carried out research on the impact of large-scale wind power grid-connected on the safe operation of the power grid, and proposed many solutions to improve the capacity of wind power consumption, but few literatures focus on the feature extraction and data dimensionality re...

AI Technical Summary

Problems solved by technology

These methods have shown their respective advantages in different fields and for different massive data types, but when dealing with massive wind power data, they have exposed their own limitations, for example: the linear method cannot extract the nonlinearity of wind power data Features; the nonlinear dimensionality reduction method based on the idea of ​​manifold learning has high requirements on the distribution density of data in high-dimensional space; while the nonlinear dimensionality reduction method based on neural network only relies on the loss function for training, it is difficult to learn completely The essential characteristics of the data; the determination of the kernel parameters by the nonlinear dimensionality reduction method based on the kernel function idea has always been a difficult point in research

And the common problem of nonlinear dimension reduction methods is how to realize the original image reconstruction of the dimension reduction results

Images