摘要
为改善具有混沌特性的风速时间序列的预测性能,提出一种混合预测方法,利用相空间重构理论实现风速时间序列的重构,通过优化周期轨道函数求取时间序列中蕴含的不稳定周期,利用前一不稳定周期的风速数据对未来风速进行预测.采用神经网络对同一风速序列再进行预测分析,将2种预测结果采用加权求和的方式进行融合,实现风速序列的混合预测,并采用混沌优化算法确定加权参数.仿真实验结果表明:混沌不稳定周期方法能够改善具有混沌特性风速序列的预测性能,混合预测方法能够进一步提高风速序列的预测效果,预测性能优于单一预测方法.
In order to improve the prediction performance of wind speed series with chaotic characteristic, a hybrid prediction method is proposed. The wind speed series are reconstructed by phase space reconstruction theory. The unstable period contained in wind speed series can be resolved by optimizing the periodic orbit function. The future data of wind speed can be predicted by the previous unstable period. The neural network is applied to predict the same wind speed series. A hybrid prediction is completed by fusing the prediction results obtained by the unstable period method and neural network. The fusion parameters can be determined by chaos optimization algorithm. Simulation results show that the unstable period method has good prediction performance, and the hybrid method can get better prediction performance than each one.
出处
《天津工业大学学报》
CAS
北大核心
2013年第5期47-50,56,共5页
Journal of Tiangong University
基金
国家自然科学基金项目(61203302)
关键词
风速时间序列
风速预测
混沌特性
不稳定周期
BP网络
混合预测
wind speed series
wind speed prediction
chaos characteristics
unstable period
BP neural network
hybrid prediction
