Research Of Wavelet Entropy Detection Methods On Power System Transient Signals

With the rapid development of power system, an increasing number of non-linear power electronics equipments has been utilized in power grid, which has become an important source of power quality problems as well as a serious threat to the stability of the power system. Meanwhile, due to human actions or natural disaster phenomena, a variety of short-circuit failure and lightning interference occurred in actual operation process, these faults need for timely detection and treatment. However, the characteristics of disturbance signals, generated by the power system faults, are easily drowned by harmonics and other background noise, which makes it difficult to identify the disturbance directly. Therefore, the realization of the signal feature extraction by filtering the noise, have a great significance for the improvement of power system stability.First of all, this paper realizes the wavelet transform and wavelet packet algorithm on the basis of analyzing the development of the feature extraction theory, and then the DTCWT is introduced to the analysis. The designed simulation results show that the DTCWT could significantly improve the existing frequency aliasing and translation sensitivity defects of the traditional wavelet transform, during the operation process. Secondly, due to the relatively low energy and small amplitude, the power system transient signal is easy to be overwhelmed by steady-state signals and system noise,therefore, the entropy theory is utilized to further process the wavelet transform results, in order to improve the simulation results. However the entropy results are still affected by the defects that the traditional Shannon entropy is only suitable for extensive system. To broaden the application of entropy method, the Tsallis entropy, which could be applied to non-extensive system, is introduced to power signal detection. After the above analysis, this paper combines the non-extensive Tsallis entropy with the DTCWT, and thus puts forward a novel Tsallis wavelet entropy approach to the feature extraction research of the transient power system signals. In addition, it designs the simulation experiments to analyze the contact and difference between Shannon wavelet entropy, Tsallis wavelet entropy and Tsallis complex wavelet entropy, and verifies the feasibility of the novel Tsallis entropy method.Finally, a typical power system model is built to produce several typical fault signals with the MATLAB/Sim Power Systems toolbox, for characteristic analysis. After that, a set of measured AC phase-to-ground signals is taken as the simulation object to do feature extraction by utilizing the proposed Tsallis complex wavelet entropy method. The simulation results show that, under the premise of maintaining a similar operation speed, the improved Tsallis entropy method increases the feature extraction accuracy of the power system transient signals significantly, compared with the traditional Shannon wavelet entropy and Tsallis wavelet entropy. Therefore it provides more effective protection for the accurate identification of power system fault types.