Electric vehicles (EVs) are both traffic loads of the transport network and electricity loads of power grid, and their travelling as well as charging behaviours will have an impact on the operating laws of the transport network and power grid. Aiming at the traditional EV prediction method of random sampling for a single EV unit, which fails to consider the operation state from the movement of EV groups, in this paper, a method for forecasting electric vehicle charging loads that leverages traffic equilibrium theory is proposed. Firstly, models for electric vehicles and road networks are constructed in both time and space dimensions, and an efficient path set for origin-destination (OD) pairs of electric vehicles is generated using the A* algorithm. Taking into account users' limited rationality, a charging station electricity pricing, which incorporates traffic flow constraints, is formulated by introducing time-of-use (TOU) pricing. Furthermore, a semi-dynamic traffic equilibrium model is developed by integrating stochastic utility theory with TOU pricing to address the traffic flow allocation. then based on the probabilistic numerical calculation method of combined state of charge (CSOC), the charging load of electric vehicles is solved. Finally, based on the data of new energy vehicles in Shanghai, the effectiveness of the proposed charging load prediction method is verified by analysing the improved 13 nodes road network and a regional road network in Shanghai, the results show that the proposed method can reduce the charging cost and alleviate the charging pressure of charging stations.