In the paper a way to build a wireless sensor network based on LoRaWAN technology is considered. The proposed approach allows to increase the intensity of data collection compared to standard implementations of LoRaWAN network and save energy efficiency of wireless sensors. The main feature of the network is the edge computing algorithm of distributed data compression. The algorithm is based on controlling prediction error on the end device side. The prediction is completed by recursive linear prediction algorithm on the server side. Higher data collection rates are achieved along with acceptable level of energy efficiency for autonomous operation of sensors and low load of the physical communication channel. The effectiveness of the proposed solution can be illustrated by the results of simulation experiments and the network of temperature sensors for intensive temperature measurement in the problem of identifying the topology of the office building heating system. The network is based on open source software, ChirpStack server, LoRaWAN end-device stack from Semtech and additional modules developed by the authors.