This paper proposes a novel online optimal mode management method for Plug-in Hybrid Vehicles(PHVs). In this work, energy consumption mapping is built for PHV model, and joint probability density functions of speed and required power are considered for each sampling distance interval since drivers'actions and traffic conditions randomly change. Then, the fuel consumption is minimized by optimizing threshold of the required power to decide which mode, Electric Vehicle or Hybrid Vehicle mode, to use for the PHV operating. The utility is demonstrated via a detailed numerical simulator. The simulation result shows that the optimization problem can be solved online.