The existing FFT-WT algorithm and FFT-BP algorithm have advantages only for the detection of certain specific harmonics in photovoltaic systems. In this paper, the FFT-WT algorithm is improved, and a DFFT-WT algorithm is proposed. The FFT-BP algorithm is improved and an improved FFT-BP model is proposed. On the basis of this, the two are combined, the threshold interval is introduced to limit the search scope of the neural network iteration, and a DFFT-WT-BP(Double FFT-WT-BP) detection algorithm is proposed. The simulation results show that the algorithm can detect complex harmonic signals composed of multiple harmonics in the actual photovoltaic network inversion system, and is more accurate and practical.