Improved Extreme Learning Machine Method for Wind Turbine Clutter Mitigation

Abstract

Due to its rapid learning capacity and well generalization performance, the Extreme Learning Machine (ELM) is creatively introduced into wind turbine clutter (WTC) mitigation for weather radar. Aiming at the difficulty of setting the number of hidden layer nodes in ELM algorithm, an improved algorithm‐‐Incremental Extreme Learning Machine (I-ELM) is proposed. First, the training samples are constructed by using the radial velocity and spectral width of the weather signal from the neighboring range bins. Then through the training of samples, the model parameters are searched and optimized according to the least square criterion. Finally, the optimized I-ELM model is utilized to recover the weather signal of the contaminated range bin. Theoretical analysis and simulation results show that the proposed algorithm can effectively suppress WTC and significantly reduce the deviation of radial velocity estimation and spectral width estimation caused by WTC contamination.