风速时间序列非线性短期预测

针对风速时间序列复杂的非线性特征,根据C-C算法确定重构参数(嵌入维数及延迟时间)并对风速重构相空间,建立径向基函数神经网络(RBF网络)及Volterra自适应预测模型对风速时间序列进行预测,以Lorenz方程数值解为例验证了两种预测方法的可行性。结果表明:RBF神经网络模型和Volterra自适应预测模型都能对实测风速时间序列进行较为准确的预测,预测误差分别在0.3和0.1 m/s内;Volterra自适应预测模型预测结果总体较RBF神经网络模型预测精度更高,且随着预测时间的增大,预测误差呈增大趋势,这与混沌存在初值敏感性的特征相符。

Nonlinear Short-Term Prediction of Wind Speed Time Series

According to the complex nonlinear characteristics of wind speed time series,the reconstruction parameters (embedded dimension and delay time) are determined according to CC algorithm,and the phase space is reconstructed by wind speed.Radial basis function neural network (RBF network) and Volterra adaptive prediction model are established.The prediction of wind speed time series is carried out.The numerical solution of Lorenz equation is taken as an example to verify the feasibility of the two prediction methods.The results show that both the RBF neural network model and the Volterra adaptive prediction model can accurately predict the measured wind speed time series,and the prediction error is controlled within 0.3 and 0.1 m/s,respectively.The Volterra adaptive prediction model predicts the overall result.The prediction accuracy of the RBF neural network model is higher,and as the prediction time increases,the prediction error increases,which is consistent with the characteristics of the initial value sensitivity of chaos.