The Super Tau-Charm Facility (STCF) is one of China's most advanced positron-electron colliders in the future, with a peak luminosity of $0.5\times10^{35}cm^{-2}s^{-1}$ and a center-of-mass energy of $2\sim7$ GeV, designed specifically to explore various physics phenomena in the $\tau$-charm energy region. Particle identification (PID) is a crucial element in physics analysis and is vital for achieving exceptional scientific performance. STCF places high demands on PID accuracy and efficiency to meet its stringent standards. In recent decades, machine learning (ML) techniques have emerged as a dominant methodology for PID in high-energy physics experiments, consistently delivering superior results. This study introduces an advanced PID software based on ML algorithms, developed for STCF to advance physics research. It includes a comprehensive global PID algorithm for charged particles, combining information from all sub-detectors, as well as a deep convolutional neural network (CNN) that utilizes Cherenkov detector inputs to effectively distinguish charged hadrons. Additionally, a CNN has been developed to differentiate neutral particles using calorimeter responses.Initial findings indicate that these PID models have demonstrated exceptional performance, significantly enhancing the scientific potential of STCF.