The Time-Dependent Orienteering Problem (TDOP) is a generalization of the Orienteering Problem where graph weights vary with time. It has many real life applications particularly associated with transport networks, in which travel time between two points depends on the moment of start. The paper presents an evolutionary algorithm with embedded local search operators and heuristic crossover, which solves TDOP. The algorithm was tested on TDOP benchmark instances and in most cases achieved optimal or near optimal results clearly outperforming other published methods.