Severe weather reduces sight distance and increases stopping distance of traveling vehicles. Traffic collisions are likely to occur when vehicles encounter traffic congestions on freeways under severe weathers. The primary objective of this study is to develop a control strategy of variable speed limit (VSL) to improve freeway safety in different types of weather conditions. The control strategy aims to reduce vehicle speeds gradually prior to congestions. A car-following model was used to simulate the behaviors of drivers in different visibility and road adhesion conditions. The time-to-collision was calculated to evaluate the collision potentials on freeway mainlines. Five typical severe weathers were tested. The simulation results showed that the VSL control strategy proposed in this study effectively reduced collision potentials under various types of severe weathers. The collision potential was reduced by 53.98% in heavy fog weather to 91.69% in heavy rain weather. The VSL strategy was compared to a previous VSL algorithm. The results showed that the strategy in this study outperformed the previous one in reducing collision potentials in most of cases.