Impedance is an important parameter of power lithium-ion batteries, which can represent battery characteristics and can also be used as an indicator for battery fault diagnosis. Since Electrochemical Impedance Spectroscopy (EIS) includes various electrode processes and information, it is more significant and worthwhile for lithium-ion batteries research. However, it is quite difficult to attain EIS online because of the nonlinear characteristics of batteries. Therefore, this paper focus on studying the nonlinear impedance characteristics of lithium-ion batteries and proposing a new method to calculate the EIS online based on Fast Fourier Transform (FFT). Data similarity analysis is used to study the influence of resting time, excitation current amplitude, bias current amplitude and the state of charge (SOC) on the impedance quantitatively. Further research finds that the amplitude of impedance, voltage and current in frequency domain all conform to the law of power function well within 6Hz, therefore, the calculation frequency scope is determined as less than 6Hz. Moreover, SOC has little influence on impedance amplitude compared with phase angle, based on this feature, the impedance amplitude is selected as calculation target. The online calculation method of impedance amplitude is proposed based on least square fitting and FFT. Furthermore, data feature extraction is used to filter the voltage and current data so that this method can be used without limited by operating conditions. Meanwhile, the factors affecting the calculation accuracy are analyzed. On one hand, the accuracy can be improved by adding Hanning window. On the other hand, shifting down impedance amplitude curve can reduce system measurement error. The data of other operating conditions and other battery are used for verification, and it is found that the relative error can be controlled within 15% approximately under all operating conditions.