As one of the key components of inverters, the health status of DC-link capacitor is closely associated with the safe operation of rail transit systems and is the potential cause of catastrophic failures. Considering the importance of condition monitoring of DC-link capacitors, capacitance is usually used as degradation parameter. However, identification error of existing methods may rise in rail transit due to limitations, such as low frequency sampling, noise interference, computational complexity. Thus, a novel Kalman filtering estimation method based on gain adjustment strategy is proposed. This method implements the capacitance estimation by iteratively filtering the pre-charging voltage. During the filtering process, the gain matrix is adjusted according to the estimation error of least square algorithm, which optimizes the weight of measurement innovation and improves the calculation accuracy of capacitance. Matlab/Simulink simulation are carried out for accuracy verification and noise immunity analysis. The results show that the maximum identification error does not exceed 0.5% with sampling intervals of 1 ms and 10 ms. Moreover, the estimation error is within 5% when the signal-to-noise ratio is greater than 50 dB. The effectiveness of the proposed method is also verified on metro vehicles with an average error below 1%. The confidence interval of the estimation error is [0.55%, 1.44%] at 95% confidence level.