With the large-scale connection of PV power generation systems to the power grid, serious harmonic pollution is inevitable. In order to effectively monitor the output current’s harmonics and inter-harmonics of a PV grid-connected system, this paper proposes a detection method based on the improved fast TLS-ESPRIT algorithm combined with the 2-order Blackman-Harris self-convolution window. First, the signal is sampled three times and the frequency is detected by the fast TLS-ESPRIT algorithm. Then the detection results are analyzed based on the simplified K-means clustering algorithm and the real harmonic components are extracted. Finally, the 2-order Blackman-Harris self-convolution window is used to interpolate the signal, and the amplitude and phase information of the signal are accurately estimated. The high-precision detection of harmonics and inter-harmonics is realized. Simulation and field data test results show that the proposed method has higher harmonic and inter-harmonic detection accuracy and stronger anti-interference ability than the traditional methods.