Abstract: Train speed control is the core of Automatic Train Operation (ATO) system, and researching effective train speed control algorithms plays an important role in improving train speed regulation performance. Therefore, this paper proposed a composite control method for train speed based on T-S type fuzzy inference soft switching: Combining fuzzy control with gain self-adjusting single neuron PID (Proportional Integral Derivative) control to form a composite control method for train speed; Based on T-S type fuzzy reasoning, implement dynamic soft switching of two types of control. Simulation tests show that this method fully utilizes the advantages of two types of train speed control and implements smooth transitions, effectively improves the speed and stability of train speed regulation response.