Software-Defined Networking is a networking paradigm that enables new approach for network applications and engineering. Its fundamental concept is based on the separation of a centralized control plane and distributed forwarding plane. It provides a global view of entire network topology, combined with the capability of programming network elements in a logically centralized manner. One of the SDN benefits is the ability to optimize network traffic more efficiently than in classical IP networks. This paper proposes a heuristic traffic engineering approach in SDN network based on multipath forwarding and switching of flows between paths. The presented approach relies on a dynamic selection of the best path in terms of network load. The path selection is made in two phases. First, an initial lowest cost choice of a path out of available k-paths is determined, then a heuristic reselection of flows according to paths load and flow properties is triggered. The developed concept has been verified using the OpenDaylight controller enhanced by the Traffic Engineering module, the Mininet network emulator, and realistic network topologies. Simulations have shown that the proposed approach provides better results over plain multi-path traffic engineering approach.