Abstract Having no information about the probabilistic laws of noises least square method has been applied to solve nonlinear filtration prob lem. It introduces the variational methods and the solution of nonline ar two-point boundary value system. Two recursive techniques for sol ving this system have been proposed and investigated. One of them based on the disturbance theory has been applied for estimation of the unk nown power system model parameters and for filtration of measurement noise simultaneously. Besides the estimates of unknown parameters and power system state, the algorithm calculates the speed and acceleration in changes of these estimates.