中文摘要：

结合相空间重构理论，针对具有混沌特性的风速时间序列提出一种基于不稳定周期的预测方法．采用互信息法计算给定时间序列的延迟时间参数，根据时间序列运行轨迹的重合度，构造不稳定周期优化函数．通过对该函数的优化计算，得到嵌入维数参数及最佳不稳定周期值．根据所得延迟时间和嵌入维数等参数对风速时间序列进行相空间重构．利用前一不稳定周期轨迹附近的数值实现对未来风速时间序列的预测分析．仿真实验结果表明，该方法能够有效提高风速时间序列的预测性能，并可实现风速序列的多步预测分析．与持续法等传统预测方法相比，当预测步长增加时，该方法具有更稳定的预测性能．

英文摘要：

Combining the phase space reconstruction theory, a novel prediction method based on the unstable period is proposed to improve the prediction performance of wind speed series with chaotic characteristics. The delay time of the known time series is determined by the mutual information method. The unstable period optimization function is constructed according to the contact ratio of the orbit of time series. The embedding dimension and the best unstable period are determined by opti mizing the unstable period function. According to the delay time and the embedding dimension, the wind speed series are reconstructed by the phase space reconstruction theory. The future data is pre dicted by the points near the former unstable period. Simulation results show that the method can en hance the prediction performance of wind speed series, and can also be applied to multistep predic tion of wind speed series. Compared with the traditional prediction methods, the prediction perform ance of the method is more stable when prediction step is increased.