Approximate entropy and its application to fault detection and identification in power swing

Most of the signals in power system are typically non-stationary signal with time-varied characteristic. By researching on Approximate Entropy (ApEn), a new technology is introduced in the analysis of non-stationary power signals because ApEn can describe the disorder or irregularity of signals. The application to ideal power signals analysis with ApEn and the comparison between ApEn and Shannon Entropy confirm the predominance of ApEn in some part of power signal analysis, so it provides an effective algorithm for power signal analysis. Take the characteristic of power swing signal into account, this paper introduces ApEn as a tool to analyze the fault identification during swing conditions in power protection and fast Wavelet Transform is introduced as a pre-process algorithm for noise reducing and high-frequency abstraction. Utilizing the above method, simulations and practical tests are done and it proves that ApEn can well distinguish two swing signals with different faults even under the condition of short time-series, small magnitude and so on. Thereby, it proves to be an effective algorithm for fault identification during power swings. Moreover, the prospect of approximate entropy´s application to power fault diagnosis has been forecasted.