中文摘要：

为提高风速序列预测的准确性,在双承接层Elman网络的基础上提出了迟滞Elman预测网络.网络具有输入层、隐层、隐层承接层、输出层以及输出承接层5层结构,并在隐层承接层和输出承接层的单元中增加了迟滞激励响应函数,从而将迟滞特性引入到Elman网络中,以提高网络处理连续信息的能力.选择梯度下降方法作为网络的学习算法,训练网络的权值及迟滞参数,利用该预测网络实现了风速序列的多步预测分析.仿真实验结果表明：迟滞特性的引入能够减小预测结果的随机波动性,有利于提高预测结果的可靠性,与现有预测方法相比,迟滞Elman网络的平均预测误差能够减小8%以上,整体预测性能以及波动较强的局部预测性能都能得到显著提高.

英文摘要：

In order to improve the prediction accuracy of the wind speed series, a hysteretic Elman network is proposed based on the Elman network with two context layers. The hysteretic Elman network is composed of input layer, hidden layer, context layer of the hidden layer, output layer and context layer of the output layer. The hysteretic characteristic is brought into the network by adding the activation functions into the two context layers, which can enhance the information processing ability for the continuous information. The gradient descent method is used to train the weights and the hysteretic parameters of the network. Muhi-step ahead prediction of the wind speed series can be performed by the method. Simulation results show that the hysteretie characteristic can restrain the random fluctuation of the prediction results, and enhance the reliability of the prediction result. Compared with the conventional methods, hysteretic Elman network can get better prediction performance on not only the global trend but also the local region with the sharp fluctuation, and its average prediction error is reduced by more than 8%.