In this paper is proposed a solution to the problem of low efficiency of bandwidth usage of physical channels. In the beginning adducted a necessity to implement proposed solutions at transport network. The main problem of transport networks is low level of utilization of optical channels resources. As a result it is often very expensive to exploit such a network due to high capital and operational costs. The method of distributed multi-flow traffic engineering in software-defined networks based on means and mechanisms of flow processing by Openflow switches has been developed. This method ensures optimal dynamic redistribution of physical channel bandwidth among several logical flows in the network, thus it achieves more than 90% of bandwidth utilization. The method uses sophisticated methods and algorithms of traffic engineering. Method using Hurst index allows evaluation of future network state and redistribution of logical flows over a different physical channels while supporting utilization efficiency of these channels above 90%. The method of load balancing uses as a basis priorities of physical branches of one logical channel. Thus this method allows the network to adapt to dynamically-changing incoming traffic, based on mechanism of prognostication the traffic rate using Hurts index. As a result algorithm of the method consists of two stages that ensure effective traffic redistribution: traffic prognostication and load balancing. Searching for optimal physical routes for one logical channel is based on Gallaher multi-path data transfer model. The method of calculating the efficiency of utilization of provided bandwidth for a particular physical channel as the basis for the mechanisms of redistribution of the total bandwidth between various logical channels is developed. Using mechanism of redistribution of flows into different physical channels controller can unload and set it to stand-by mode those segments of the network that are low-utilized. To verify the effectiveness of the proposed method of load balancing simulation model was created based on distributed system of servers and software models of SDN components developed using C++ language. The result of modeling confirmed the expediency of introduction the method on a real transport network.