Accurate and On-Line Timing using Recurrent Wavelet Neural Network for Power Systems

Global Positioning System (GPS) based time reference provides inexpensive but accurate timing and synchronization capability and meets requirements in power system fault location and monitoring. Global Navigation Satellite System (GNSS) signals from approximately eleven thousand miles above the earth provide a valuable fourth time. This universally dimension-precise available service is improved and changed the way many businesses accurately track, mange, and synchronize their operations. Precise timing, made available through GNSS, is playing an ever increasingly role in the expansion of time-critical applications on a global basis. Neural network have been established as a general approximation tool for fitting nonlinear models for input-output data. On the other hand, the recently introduced wavelet decomposition emerges as a powerful tool for functional approximation. In this papers, a Recurrent Wavelet Neural Network (RWNN) and its learning algorithm is presented for prediction of GPS receivers timing errors. This method is well suited for real time improving of GPS receivers timing accuracy. The proposed NN is implemented by a L1 GPS receiver. The experimental test results on the collected real data are presented and discussed in this paper. The experimental results emphasize that GPS timing error RMS can reduce from 300nsec and 200nsec to less than 183nsec and 51nsec by using RWNN prediction, before and after SA, respectively. Also, it is shown that performance RWNN is better than single WNN and single RNN.

کلیدواژه ها:

Prediction ، Timing Error ، GPS ، RWNN ، Power Systems