Removing decaying DC component in fault currents via a new modify discrete Fourier algorithm

In protection relaying schemes, the discrete Fourier transform (DFT) is the most widely used algorithm for computing the fundamental frequency phasor. When the measurements only contain fundamental frequency and integer harmonic frequency components, the conventional DFT only needs "one-cycle samples" to calculate the fundamental frequency phasor. However, the voltage and current signals usually contain the decaying dc component during the fault interval. These decaying components involved in measurements will postpone the convergent speed of the DFT algorithm. In most cases, the conventional DFT will postpone about 4 cycles (or more) to calculate the accurate fundamental frequency phasor. The slow convergence will reduce the accuracy and response time of the following fault locator or other installations in the relaying schemes. In order to overcome the above problems, many works have been proposed, such as the digital mimic filtering and the smart DFT techniques. However, the high pass filter characteristic of mimic filter will cause the computation becomes noise sensitivity. Meanwhile, if the decaying dc does not contain in the measurements. The smart DFT computation will go divergently. Therefore, the applications of the above techniques will need the auxiliary components to prevent the above problems. This work presents a new modify DFT to remove the decaying dc component and calculate the accurate fundamental frequency component. The proposed algorithm only needs "one cycle plus one samples" for computation. Thus the proposed algorithm is faster than the above one. The algorithm can sustain the noise contains in the measurements. In addition, when the decaying dc component disappears, the computation result of the algorithm will be the same as the conventional DFT. Therefore, the proposed new DFT algorithm is also robust. Some simulation results in the paper are presented to demonstrate the effect of proposed new algorithm.