迟滞混沌神经网络及其对风速序列短期预测的研究

项目名称： 迟滞混沌神经网络及其对风速序列短期预测的研究

项目编号： No.61203302

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 修春波

作者单位： 天津工业大学

项目金额： 24万元

中文摘要： 本课题利用迟滞混沌算子单元，构造一种同时具有迟滞和混沌两种非线性特性的神经网络，通过深入研究网络的动力学特性，采用充分利用网络非线性特性的方式，挖掘网络的信息处理能力，并应用于风速时间序列的短期预测研究中。针对实际被预测系统的复杂性，从机理上区别于现有的静态建模预测方法，实现一种动态建模预测方法。即通过不断改变网络各单元的动力学特性来调节网络的动力学行为，使网络的动力学特性逐渐逼近被预测系统的动力学特性，并保持与之一致变化。利用迟滞特性提高网络对有用信息的利用率，以此增强网络的记忆能力，从而提高网络的泛化能力，达到提高网络预测精度的目的。该课题的研究一方面丰富现有神经网络的种类，从理论上促进神经网络及相关学科的发展；另一方面提供一种时间序列预测的新思想，从实现机理上克服静态建模与预测的缺点。另外，本课题对风速的预测研究将为风能的开发和利用起到积极的促进作用，符合新能源与低碳理念的要求。

中文关键词： 风速序列；迟滞；混沌；神经网络；控制

英文摘要： A novel neural network which has two kinds of nonlinear characteristics, hysteresis and chaos, will be constructed by hysteretic chaotic operator units. Its dynamic characteristic will be investigated, and its application potential on information processing will be exploited by utilizing its nonlinear characteristics. Subsequently, we will use the network to resolve short-term wind speed forecasting. Based on the complexity of predicted system, a novel dynamic prediction method, which is different from the static modeling prediction method, will be proposed. In other words, the dynamic characteristic of the network can be changed by adjusting the characteristic parameters of operator units to approach and consist with that of the predicted system. Hysteretic characteristic can help the network to strengthen the information exploitability, increase the memory ability, improve the capacity of generalization, and enhance forecasting accuracy. On the one hand, the project can increase the varieties of neural network, and can promote the development of the neural network and related subjects. On the other hand, the project can give a novel dynamic prediction method for time series, which can overcome the shortcomings of static modeling and prediction method. In addition, the wind speed prediction can promote the deve

英文关键词： wind speed series；hysteresis；chaos；neural network；control