Abstract— In power systems, protective relays must filter their inputs to remove undesirable quantities and retain signal quantities of interest. This job must be performed accurate and fast. A new method for filtering the undesirable components such as DC and harmonic components associated with the fundamental system signals. The method is s based on a dynamic filtering algorithm. The filtering algorithm has many advantages over some other classical methods. It can be used as dynamic on-line filter without the need of parameters readjusting as in the case of classic filters. The proposed filter is tested using different signals. Effects of number of samples and sampling window size are discussed. Results obtained are presented and discussed to show the algorithm capabilities. Keywords — Protection, DC-offset, Dynamic Filter, Estimation. I. I NTRODUCTION N power system protection, the fundamental frequency voltage or current signals must be decomposed by relays to enable fast response. One of important protection applications is distance protection. In such protection system, to locate the fault accurately and fast the fundamental voltage and current signals must be extracted on-line. Therefore digital filtering in distance relays includes some form of signal parameter estimation algorithm. The resistance-inductance behavior of the power system produces dc- offset components in the fault current which are exponentially decaying. It is known that distance relays have the tendency to overreach in the presence of dc- offset component in the faulted current waveform. A number of techniques has been proposed to remove the dc-offsets from the fault currents. Digital filters, which eliminate the effect of an exponentially decaying component, were proposed in many references. Some of these filters are based on static and dynamic techniques [1], [2], [3]. Sachdev et-al [1] proposed the application of the well known least error square estimation technique. For decomposing the fault current and removing the dc off-set components [1]. The least error squares is an accurate estimation technique but in case of the data set is contaminated with bad measurements, the solution may not be accurate unless extra filters are used for detecting the bad measurements. A digital mimic filter for the removal of DC-offset current waveforms is presented in reference 2. The proposed filter in this reference is used in conjunction with other classic estimators such as least error square to detect the whole system signals components for relaying purposes. Reference 3 proposed the use of Kalman filtering