Dynamic frequency scanning interferometry (DFSI) offers the advantage of drift immunity in distance measurements, making it widely used in industrial applications. However, the nonlinear scanning characteristics of lasers can induce non-stationarity in interference signals, complicating the phase extraction process and decreasing the accuracy of dynamic distance measurement. This study investigates the phase extraction of non-stationary signals with multiple frequency components and proposes a second-order group delay error compensation algorithm. By utilizing complex Morlet wavelets and second-order time-varying signals to construct a group delay model, this algorithm enables precise extraction of non-stationary phases within specific frequency bands, thereby mitigating phase extraction errors in DFSI systems. Experimental results showcase the effectiveness of the algorithm in reducing the standard deviation of distance measurements from 20.26 μm to 11.47 μm by mitigating the nonlinearity. Furthermore, dynamic experiments validate the stability of the proposed algorithm in detecting vibrating and uniformly moving targets.