Abstract: Applied beamspace root-MUSIC, with signal spaces as the models, to do harmonic detection to overcome the drawbacks of the traditional methods. Firstly, Gerschgorin disk criterion was employed to detect the source number, then the data acceptable to the array was transformed to get the data of the beamspace, which was further decomposed to noise subspace and signal subspace. Next, a polynomial was constructed with the use of the orthogonality of these two subspaces. By solving this polynomial, the estimate frequency value of the signal was attained. Lastly, with the extension of the PRONY method, the amplitude of the signal was achieved. Simulation experiments and real data analysis show that this algorithm in power system harmonic detection is feasible.