To build models of a multi-level process for collecting and processing data from a telecommunication system for distributed data acquisition and thermoelectric control, including separately allocated objects, the most effective is use of wireless sensor ad-hoc networks (Wireless Ubiquitous Sensor Networks, WUSN). A method is proposed to adjust the performance of WUSN with adaptation of the link layer based on neuro-fuzzy approach, which performs estimation of the dynamic behavior of a communication channel, as well as the integrity of the received data, and controls information packets retransmissions. A comparison was made of the performance of the WUSN network based on the well known CSMA/SA link layer protocol, as well as a similar network with link layer adaptation. It has been established that the proposed link layer protocol provides the least quantity of missed packets at a similar transmission speed and average delay.