In this paper, a frequency domain modelling method which estimates power system harmonics and interharmonic components occurring in frequency domain because of non-linear loads is introduced. In this method, the difference between original signal and frequency domain modelled signal is minimized and solved as a nonlinear least squares problem. To solve this problem, genetic algorithm optimization technic is used to find frequency locations. Then, the estimated frequency values are used in the model equations to determine amplitude and phase components of the estimated frequencies. Eliminating spectral leakage ensues from windowing and more accurate detecting amplitudes and phases of power system harmonics are ensured by the proposed method. The proposed method is applied to synthetic data, consisting harmonic signal components, created on MATLAB, and tested the accuracy of the method. After then, the proposed method are applied to field voltage data, feeding electric arc furnaces in Iskenderun region containing harmonics, taken from Turkey Electricity Transmission System and it has been shown that more accurate results provided according to inverse Fourier Transform technique.