The invention provides a method for identifying power generator unloaded characteristic linearity range on the basis of linear condensation degree. The method comprises following steps: (1) through a power generator unloaded characteristic test, a set of power generator unloaded characteristic test data is obtained: X={Ii, Ui}, i=1,2,...,n; wherein Ui represents measured generator terminal voltage value, and Ii represents measured field current value; (2) linear condensation degree sequence is calculated according to the following formula: D(X)={di}, i=1,2,...n, wherein di represents linear condensation degree constantly corresponding to data {Ii, Ui}: di=(U-Ui)Ii/[(I-I)Ui], i=1,2,...,n-1; the values of linear condensation degree are compared; data from 0.9 to 1.1 is selected and correspondent original test data is the linearity range data; the linearity range slope is the slope of a curve where the corresponding linear condensation degree is closest to 1: K=(U-U)/(I-I), and |dj-1|=min(|di-1|), i, j=1,2,...,n-1. According to the invention, data length and slope of linearity range are identified through calculation of linear condensation degree of power generator unloaded characteristic; errors and inconsistency due to artificial identification are avoided; the method has strong identification capability for test data having interference. The method is suitable for power generator unloaded characteristic test and power generator parameter calculation.