基于相空间重构与模糊神经网络的海温垂直分布预测模型  

Application of Phase Space Reconstruction and ANFIS Model for Forecasting Sea Temperature Vertical Distribution

作　　者：董兆俊[1] 滕军[1] 张恒新[2] 张芳苒[1] 赵建宇[1] 

机构地区：[1]中国人民解放军61741部队 [2]中国人民解放军93886部队气象中心

出　　处：《海洋科学进展》2010年第2期158-162,共5页Advances in Marine Science

摘　　要：在海温预报中引入混沌理论，将相空间重构与模糊神经网络相结合，提出了海温垂直建模预测模型。通过相空间重构，把海温时间序列拓展为多维序列，而多维序列包含着各态历经的信息，从而挖掘出了丰富的海温变化空间的信息，有利于模糊神经网络的训练。利用建立好的模糊神经网络模型，较好地对海温的垂直结构进行了建模、训练和预测。实际的预测结果表明，该模型预报精度较高，超前1～5个月的预测值的相对误差均控制在10％以内，预测结果可以为业务工作提供一定的参考与借鉴。The chaos theory is introduced and employed in the sea temperature forecasting, and the phase space reconstruction and the ANFIS model are combined to establish a model for forecasting vertical distribution of sea temperature. Based on the phase reconstruction, the time series of sea temperature is expand- ed into the multivariate time series including ergodic informations, to obtain more abundant informations of sea temperature variability in favor of ANFIS training. The ANFIS model can work well in the stages for establishing, training and forecasting. The method evidently improves the precision in forecasting the sea temperature vertical distribution with the relative errors lower than 1% in the 1～ 5 month forecast. The results may provide the reference to the operational work.

关 键 词：相空间重构 模糊神经网络 海温垂直分布 预测模型 

分 类 号：P731.11[天文地球—海洋科学]