Coherency group identification is an integral constituent part of the wider field of reduction techniques in power systems. It consists of separating the machines in the system into groups that feature similar behavior. This paper presents a coherency identification algorithm for dynamic studies. The algorithm combines both modal and time domain techniques in an effort to combine the merits of both approaches. Its outcome is a suggested optimal number of clusters alongside the clustering itself. Tests have been conducted on a sample power system of 39 buses and its validity has been demonstrated.