摘要
针对风速序列非线性和非平稳性的随机性特点,提出了基于小波过程神经元网络的短期风速预测方法。首先利用相空间重构理论,计算出风速时序的最佳嵌入维数作为网络的输入层节点数,根据小波神经网络的经验公式来选取网络隐含层的节点数初始值,通过调整参数使网络误差达到最小值,得到合适的隐层节点数,并给出相应的学习算法。算例仿真结果表明所提预测方法的可行性,运用本方法与时序ARMA模型对比,其预测结果的精度明显提高。
According to the non-stationary random and nonlinear characteristics of wind speed, a short-term wind speed forecast based on wavelet process neural network is proposed in this paper. First, by using the theory of phase space reconstruction, the optimal embedding dimension of wind speed sequence is calculated, and the network input node number is determined. Then, by using empirical formula, the initial nodes number of the wavelet neural net- work hidden layer is selected. By adjusting the parameters the minimum network error can be reached, the network proper hidden nodes are gotten. The corresponding learning algorithm is set. The simulation result shows that the proposed method is feasible, and using the proposed method can get better accuracy of forecast result compared with timing ARMA model.
出处
《电子测量与仪器学报》
CSCD
2013年第10期944-950,共7页
Journal of Electronic Measurement and Instrumentation
关键词
过程神经网络
小波
相空间重构
process neural network
wavelet
phase-space reconstruction
