Study on signal processing of Doppler wind LiDAR at wavelength of 1550 nm based on variational mode decomposition

Coherent Doppler wind LiDAR retrieves wind speed by measuring the Doppler shift of aerosol backscattered light, offering advantages such as clear-air detection, high spatial-temporal resolution, and non-contact measurement. However, the echo signal is extremely weak and decays exponentially with range, compounded by detector noise, background light, and turbulence interference. Consequently, the signal processing stage directly determines the reliability of frequency shift extraction, wind speed accuracy, effective detection range, and data acquisition rate. This study introduces the particle swarm optimization algorithm, using sample entropy as a fitness evaluation metric to adaptively optimize the optimal decomposition level K and penalty factor α , enabling more precise parameter selection and overcoming the limitations of empirical values. Multi-strategy denoising is applied to lidar signals, and the performance advantages of each algorithm are compared. Both simulation tests and field measurements validate the effectiveness of the proposed method: in simulated experiments, the signal-to-noise ratio (SNR) is improved by 10.07 dB, while in real signal processing, SNR is enhanced by 10 dB at most compared to pure wavelet threshold denoising, rising from 10.07 dB to 21.34 dB, which fully confirming its effectiveness and feasibility for lidar signal denoising. This framework requires no prior model, can process non-stationary signals in real time, and provides a practical and efficient engineered solution for high-precision wind field measurement under weak echo conditions.