This paper examines the problem of system identification in the presence of
unknown disturbances. Specifically, we consider the case where the disturb
ance effect on the output is modeled explicitly and various ways to separat
e it from the system disturbance-free dynamics,Available for identification
are the excitation signals and the resultant responses, which may be corru
pted by unknown and possibly dominating disturbances. We assume that no act
ual measurement of the disturbances themselves is available for identificat
ion. The case where the disturbance profile is unknown and mag. be quite co
mplicated, but its period is known, is considered first, Next, we consider
the case where the disturbance frequencies are known only approximately, an
d a procedure is applied to iteratively refine the estimation of the distur
bance frequencies fur successful system identification. Finally, we examine
the situation where the unknown disturbance is nonperiodic. The identifica
tion produces results that can be used far the synthesis of feedforward and
feedback control to cancel the disturbance effect. Both simulation and exp
erimental results will be used for illustration. The simulation is carried
out with a model of a communications satellite, and the experimental result
s are obtained from a flexible truss structure. A companion paper addresses
the system identification problem where the disturbance effect is modeled
implicitly.