The tight sand-conglomerate reservoir of Shahezi Formation in the northern Songliao Basin is a continental rifted lacustrine basin deposit with deep burial depth, rapid sedimentary facies change and complex lithology combination. It is difficult to accurately reflect the geological characteristics by using traditional waveform clustering technology or other single attribute for seismic facies identification. In this paper, classification criteria were established firstly by analyzing the reflection characteristics and seismic facies patterns of different sedimentary facies zones and different lithologic combinations of typical Wells. At the same time, sensitive seismic attributes of thick sand-conglomerate were extracted and optimized for the target layer. The KNN (K-nearest neighbor) algorithm was introduced for the first time to establish a sample classifier according to classification criteria, effectively supervising seismic multi-attribute fusion. Finally, the multi-attribute fusion plan was obtained, which can accurately identify the development area of thick sand-conglomerate area. On this basis, combined with the sand-ground ratio of single well, single well facies and seismic facies, the comprehensive analysis gives sedimentary significance, and further analyzes the planar distribution characteristics of sedimentary facies. The results show that sequence 4–3 of Shahezi Formation in SS1 well area in northern Songliao Basin mainly develops fan delta, braided river delta and shoreline shallow lake sedimentary facies. The application of this technology effectively avoids the unreliability caused by a single artificial division of the range, improves the prediction accuracy, and has an important supporting significance for the exploration breakthrough of deep tight gas.