On-line statistical identification of the nuclear reactor processes has been considered theoretically. These techniques are not time consuming and enable decreasing the uncorrelated noise effect in the overall measured noise if only the output signal of the system is accessible for the measurement.

The assumption has been made that process dynamics are approximated by single time-constant transfer functions. The postulated techniques have been based on proper filtering of the process spectral power. The auto-correlation and cross-correlation functions for the zero time lag have been applied to assignment of the spectral power. On the basis of examples, it has been shown that the proposed identification techniques enable: 1) determination of the process dynamics in a unique way by making use of the correlation functions for the zero time lag even when the process dynamics is characterized by more than one time constant; and 2) a decrease in the contribution of the uncorrelated noise in the overall measured noise when the cross-correlation technique cannot be used.