In various data transmission systems, often there are situations when one of the nodes is less productive or more overloaded than the others. Where this node is the only one on the way to data recipients, there is a strong slowdown in data transmission from sources to recipients, and some packets begin to be lost when a certain high load level is reached. The described situations are characterized by a load on the channel without strong variations in intensity and without the threat of a complete shutdown of the system, where the use of coding can have high efficiency. As a method of protection, we chose the reaction method, based on the use of correction codes that are able to restore lost packets due to the occurrence of congestion in the networks. At the same time, due to the variability in the intensity of the data flow in the channel in real systems, in order to reduce the overhead costs when using encoding and generally increase the data transfer rate, it is necessary to choose the data transfer mode (in some cases, encoding will not be more effective than a simple mechanism for re-requesting lost packets) and to determine the code parameters depending on the average probability of packet loss in the channel, which will be sufficient to restore all lost packets. In both cases, it was decided to choose the parameters of the correction code by predicting certain characteristics of the data transmission channel. These characteristics are the average delay between two packets in the channel and the queue load p. After comparing the simulation results with the optimal indicators for a similar system, the high efficiency of the method under consideration was revealed, which suggests that it will provide significant resistance to moderate loads in the channel.