As the existing estimation equivalence method for online safety analysis of power systems is of low accu-racy and weak convergence,a refined model of equivalent system,which comprises the equivalent generator andthe composite load,is proposed to improve the accuracy of dynamic equivalence,and a parameter identificationstrategy based on differential evolution (DE) algorithm is presented,Moreover,in order to overcome the prematureconvergence of DE,two differential evolution swarms with different variations are constructed,which help to in-crease the diversity of swarms and improve the convergence of DE algorithm through parallel evolution and regularmessage exchange.Simulated results show that (1) the improved DE algorithm with parallel swarms is effective inthe parameter identification of equivalent systems and is of simplicity and rapid convergence; (2) it helps to identi-fy parameters with high accuracy and strong robustness; (3) the obtained equivalent system model accords wellwith practical power systems and effectively preserves the main dynamic characteristics of the external system; and(4) the new estimation equivalence method is applicable for the online equivalent simplification of large- scale ex-ternal power systems.