Abstract: Severe convective winds (also known as thunderstorm gusts) represent a frequent hazardous convective weather phenomena in the Yangtze Reiver Delta region of China. Known for their sudden onset, high locality, and destructive potential, these events pose serious threats to economic activities public safety. However, operational forecasting of such winds remains challenging, with generally low prediction scores, indicating an urgent need for more regionally adaptive potential forecast methods. Using documented cases of level-8 or above severe convective winds (≥17.2 m/s) in the Yangtze River Delta from 2016 to 2022, this study systematically examined the environmental conditions associated with these severe wind events with a particular focus on comparing the performance of conventional physical parameters and newly developed predictors within a random forest framework. The results show that individual thermodynamic or dynamic parameters have limited discriminative power, whereas the newly constructed parameters, designed based on physical mechanisms, rank highly in feature importance, demonstrating the value of parameter integration. The random forest model performs robustly even with a limited sample size, achieving AUC values of 0.792 on the conventional test set (2016–2021) and 0.744 on an independent validation et (2022). Although the overall precision is relatively low, the high recall rate indicates a strong ability to capture severe wind events. The study identifies cold pool intensity, mid-low layer lapse rates, vertical wind shear, relative humidity as key factors influencing severe convective wind occurrence in the Yangtze River Delta. These findings suggest that the new parameters LR04SHIP,LR16CAPE and RH02SHIP proposed in this study enhance the interpretability and practical utility of convective wind forecasting, thereby, providing a scientific basis for improving region-specific prediction indicators and early warning capabilities for severe convective weather.