Strong tracking filter has some problems in terms of power quality disturbances detection, such as inadequate accuracy in strong nonlinear system, calculation complexity in high dimensional filter model and poor filtering performance. This paper combines strong tracking filter with unscented Kalman filter and proposes a new method which is based on modified strong tracking Kalman filter (MSTUKF). When the sudden changes of the state variables occur, it derives the sufficient conditions for the establishment of the MSTUKF by the suboptimal fading factor which adaptively adjusts the weights of process noise covariance matrix on the condition that filter residual sequence is orthogonal at different time. The proposed method improves the filter estimation accuracy of the filter without calculating the Jacobian matrix when compared with the traditional STF method. Besides, it just needs one UT transformation. Therefore, the computational complexity is reduced, what’s more, it retains the strong robustness of STF under model mismatch. Simulation results show that the proposed method is more rapid and has a higher accuracy detection of mutation starting and ending time of voltage sag, transient pulse and transient harmonic signal, tracking the mutation and mutant amplitude phase, which verifies the modified MSTUKF is a good solution for power quality disturbance detection. This work is supported by China Postdoctoral Science Foundation (No. 2015T80961) and Natural Science Foundation of Chongqing (No. cstc2016jcyjA0214).