Geodetic measurements are commonly used in displacement analysis to determine the absolute values of displacements of points of interest. In order to properly determine the displacement values, it is necessary to correctly identify a subgroup of mutually stable points constituting a reference system. The complexity of this task depends on the spatial size of the network, the timespan of measurements and geological conditions affecting the type of changes in the location of points. As a consequence of the abovementioned factors, the task of stable identification in a longer timespan for a subgroup of points may produce equivocal results. In particular, it is likely that alternative subgroups of reference points meeting the mutual stability criteria will be selected, sometimes without common reference points. The proposed method of reference system identification utilises optimisation algorithms. Two such algorithms were tested, i.e., simulated annealing (SA) and Hooke-Jeeves (HJ) method. Two numerical examples were used to test the proposed method. Although in the first example both methods delivered a positive result, the second example showed the superiority of the SA method over the HJ. The proposed method can be considered a tool supporting the person analysing and making calculations in reaching the ultimate decision on reference points.