Abstract: In order to achieve the goal of "carbon peaking and carbon neutralization", a large number of grid-connected inverters are connected to the distribution network. A complex interaction with the weak grid is formed, and the problem of harmonic degradation is prominent. Accurate acquisition of dq impedance of grid-connected inverters is one of the keys to analysis harmonic degradation. The existing impedance measurement methods are not suitable since the operating points of grid-connected inverters maybe time-varying and the calculation results of windowed fast Fourier transform (FFT) in frequency domain is distinct. Aiming at these defects, this paper presented a gray-box identification method, which took the white-box modeling results as the goal and the black-box algorithm as the solution. Based on the time-domain data of voltage and current of ports, the dq impedance at different operating points was identified online. Firstly, multiple sinusoidal voltage disturbances were injected into point of common coupling of the inverters in series to obtain the data needed for the identification model. Secondly, the gray-box identification model was established, and the dq impedance of the grid-connected inverters was identified online through the recursive least square algorithm. Finally, a hardware in the loop real-time simulation platform was built based on RT-LAB to verify the effectiveness of the presented method. Moreover, based on the impedance identification results, the Nyquist curve of characteristic function of the return rate matrix was drawn, and the phase angle margin of the system under different operating points was solved, so as to realize the application of method presented in this paper in the online assessment of harmonic degradation risk.