With the continuous growth of wind power access capacity, the impact of intermittent and volatile wind power generation on the grid is becoming more and more obvious, so the research of wind power prediction has been widely concerned. The dynamic behavior of wind power time series is the external performance of complex nonlinear and multi-scale phenomena. Before choosing an appropriate prediction model, it is of great significance to analyze the characteristics of wind power time series for wind power grid system. At the same time, different sampling time scales of wind power also have an important impact on its dynamic characteristics. In this study, the chaotic dynamics behaviour of wind power time series at different time scales is discussed. The research methods include stationarity and white noise judgment, power spectral density analysis, autocorrelation function analysis, probability distribution, the Hurst index, 0–1 test algorithm for chaos, correlation dimension, maximum Lyapunov exponent, Kolmogorov entropy, recurrence plot, and information entropy method are adopted to study the chaotic dynamic behavior of wind power time series at different time scales. The actual wind power data of a wind farm at different time scales are taken as the research object for the case study. The case study results draw the following conclusions: (a) The wind power time series is non-stationary and non-white noise. (b) The direct-current and low frequency components parts store the main energy. (c) In the long-term, wind power is unpredictable. (d) The output level of wind power depends on the time scales. (e) The wind power time series obeys the fractal Brownian motion. (f) Wind power time series has chaotic characteristic. (g) Wind power time series has fractal characteristics. (h) As the time scales changes, so does the maximum prediction horizon. (i) With the decrease of time scales, the information loss rate has also declined. (j) Wind power at large time scales is more chaotic. The research results of this study have certain theoretical value and practical significance for grasping the fluctuation law of wind power and improving the prediction accuracy of wind power.

随着风电接入能力的不断增长，间歇性和波动性风力发电对电网的影响越来越明显，因此风电功率预测研究受到广泛关注。风电时间序列的动态行为是复杂非线性多尺度现象的外在表现。在选择合适的预测模型之前，分析风电系统的风电时间序列特征具有重要意义。同时，不同的风电采样时间尺度对其动态特性也有重要影响。本研究讨论了不同时间尺度下风电时间序列的混沌动力学行为。研究方法包括平稳性和白噪声判断、功率谱密度分析、采用自相关函数分析、概率分布、Hurst指数、混沌0-1检验算法、相关维数、最大Lyapunov指数、Kolmogorov熵、递推图和信息熵方法研究风电时间序列的混沌动力行为不同的时间尺度。以某风电场在不同时间尺度的实际风电数据作为案例研究的研究对象。案例研究结果得出以下结论： (a) 风电时间序列为非平稳非白噪声。(b) 直流和低频分量部分存储主要能量。(c) 从长远来看，风电是不可预测的。(d) 风电输出水平取决于时间尺度。(e) 风力时间序列服从分形布朗运动。(f) 风电时间序列具有混沌特性。(g) 风电时间序列具有分形特征。(h) 随着时间尺度的变化，最大预测范围也会发生变化。(i) 随着时间尺度的减小，信息丢失率也有所下降。(j) 大时间尺度的风电更加混乱。本研究的研究成果对于掌握风电功率波动规律，提高风电功率预测精度具有一定的理论价值和实际意义。