提出了多变量混沌时间序列相空间延迟重构中延迟时间间隔和嵌入维数的选取方法 ,给出了多变量混沌时间序列的局部平均预测法 ,局部线性预测法和BP神经网络预测法等 3种非线性预测方法 .通过Lorenz系统的仿真计算表明 ,无论用 3种非线性预测方法中的哪一种 ,多变量混沌时间序列要比单变量混沌时间序列的预测误差小得多 ,即使前者的数据长度只有后者的一半 ,前者的预测误差也要小很多 .另外从预测误差最小的角度验证了多变量混沌时间序列相空间延迟重构中延迟时间间隔和嵌入维数选取方法的有效性 A method of choosing delay time interval and embedding dimension in phase space delay reconstruction of multivariate chaotic time series is proposed. Local average prediction method, local linear prediction method and BP neural network prediction method are given for multivariate chaotic time series.3 Non-linear prediction method. The simulation results of Lorenz system show that the multivariable chaotic time series is much less than the univariate chaotic time series prediction error irrespective of the three kinds of nonlinear prediction methods, The data length is only half of the latter, and the former also has a much smaller prediction error.In addition, the validity of the method for selecting the delay time interval and embedding dimension in phase-space delay reconstruction of multivariable chaotic time series is verified from the perspective of the smallest prediction error