Method for Analysis of Non-stationarity of Fluctuating Winds Based on Revised Local Recurrence Rate

In order to overcome the misestimate of the non-stationarity of different signals by the recurrence trend (RT), the mutual information function and false nearest neighbors are employed to determine the time delay and minimum embedding dimension of the recurrence quantification analysis (RQA), respectively. Then a novel index, i.e., the standard deviation of normalized local recurrence rate (SDNLRR), which is based on the RQA, was proposed to quantify the non-stationary degree of the signals. Utilizing the SDNLRR, the non-stationarity of four basic signals (white noise signal, sinusoidal signal, amplitude-modulated signal and linear frequency modulation signal) and two field-observed fluctuating wind speed histories were analyzed and compared with the analysis results of RT. The results show that the proposed SDNLRR could offer a quantitative comparison of the non-stationarity of the above six signals with a 100% accuracy. The new method eliminates the misestimates of the sinusoidal signal and the stationary fluctuating wind speed signal in RT, and hence is more accurate than the RT estimation by 33.33%.