Aiming at the problem that the existing desensitized unscented Kalman filter takes a large amount computational time to algebraically solve the gain matrix and the sensitivity weights cannot be adjusted in real-time, an adaptive fast desensitized unscented Kalman filtering algorithm is proposed. Based on the desensitized control technology, the algorithm redefines the sensitivity weight matrix of the desensitized unscented Kalman filter, and the sensitivity of the state estimation error for uncertain parameters is introduced to the cost function of the filter. The analytical solution of the filter gain matrix is obtained by minimizing the cost function, which reduces the computational complexity and time of the filter. At the same time, based on the orthogonal principle of measurement residuals, the adaptive fading factor of sensitivity weights is designed to realize the real-time adjustment of sensitivity weights in the filtering process. Numerical simulation results of a typical example show that the proposed adaptive fast desensitized unscented Kalman filtering algorithm can improve the computational efficiency, adjust the sensitivity weights in real-time and effectively reduce the adverse impact of uncertain parameters on state estimation. Compared with the traditional desensitized unscented Kalman filtering algorithm, the state estimation error and computational time of the proposed algorithm are reduced by 19.5% and 99.9% respectively.