引入混沌时间序列理论有效解决非线性混沌时间序列的预报问题.表征混沌特性的分维数D和最大Lya-punov指数1λ计算结果表明变形时间序列具有混沌特性;神经网络克服了模型必须是基本观测数据的线性和非线性组合的局限,具有很强的自适应性和记忆功能.结合混沌特性将一维时间序列重构为多维相空间,优化RBF神经网络结构,建立混沌时间序列神经网络预报模型.实例表明,预测值与实测值的相对误差小于6%;后期预报的变形随时间增长呈下降趋势,符合工程实际情况. The chaos time series theory is introduced to effectively solve the problem of nonlinear chaotic time series forecasting.Calculation of fractal dimension D and maximum Lya-punov exponent 1λ of the chaotic characteristic show that the deformation time series has the chaos characteristic. The neural network overcomes that the model must be a basic observation The linear and non-linear combination of data, has strong self-adaptability and memory function.Combining the chaotic characteristics of one-dimensional time series into multi-dimensional phase space, RBF neural network structure optimization, the establishment of chaotic time series neural network prediction model The example shows that the relative error between the predicted value and the measured value is less than 6%, and the deformation of the post - forecast decreases with time, which accords with the actual situation of the project.