A class of modified state-space self-tuning controllers of the MSEV (minimum state error variance) type is considered in this article. A suitable chosen structure for the proposed controller provides for tracking the time-varying reference input, and makes it possible to apply this solution to nonlinear and nonstationary plants. Starting from the changes of innovations sequence statistics, an efficient load disturbance detector is also constructed, and the estimated disturbance amplitude is used to correct the control signal, in order to eliminate the influence of disturbances. The advantage in using the proposed algorithm for nonlinear systems control, in the presence of load disturbances and stochastic disturbances of unknown statistics, is demonstrated through simulation results.