This study advances the field of low altitude wind field detection by systematically evaluating Doppler wind lidar performance against in situ balloon radiosonde under complex atmospheric conditions. We propose a novel machine learning framework for wind profile correction and the Aeolus satellite is used to verify the reliability of the algorithm further to enhance data reliability in meteorological remote sensing.