Increasing plug-in electric vehicle (PEV) charging stations are coupling the power and transportation systems tightly together. The distribution of PEV traffic flows will be constrained by and inversely affect both systems. In this paper, we study the PEV routing problem on a coupled transportation and power network from the perspective of a social coordinator. We formulate an interdisciplinary second order cone programming model that optimizes PEVs' driving paths and charging locations to minimize the system's social costs, which include driving and charging time costs of PEV drivers and power supply costs. The model employs: 1) an expanded transportation network model to explicitly describe PEVs' driving range constraints on the transportation network; 2) the AC power flow model to describe the electrical constraints of the power system. We then design an iterative column generation algorithm to efficiently solve it. We validate the proposed method on a coupled transportation and power network with distributed renewable generation. Numerical simulation results show that routing PEV traffic flows adopting the proposed strategy can effectively improve social welfare and promote renewable generation integration.