【作者】 于文静；

【导师】 田涛；

【作者基本信息】 华北电力大学（北京） ， 控制理论与控制工程， 2009， 硕士

【摘要】 传统的辨识方法多针对线性系统,对复杂非线性系统的辨识显示出诸多的不足。神经网络的非线性、并行分布处理和自学习等特性为解决这些问题提供了新的出路。本文着重对ELMAN神经网络的结构和动态记忆特性进行了研究,并验证了通过增加连接承接层和输出层之间的权值可以加快网络的学习速度。将改进的ELMAN网络应用于再热汽温的动态系统建模中,可见Elman网络在辨识线性系统方面与最小二乘都有良好效果;同时由于Elman网络具有自学习和映射任意非线性系统的能力,因而其在非线性系统辨识中也有较好的逼近能力。更多还原

【Abstract】 Most traditional system identification methods were applied to linear system,and it is still too difficult to solve the problems of non-linear system.Neural network has Characteristics such as non-linear,parallel distributed processing and self-learning which give us a new way on system identification.This paper research on the structure and dynamic memory characteristics of Elman neural network and increased the learning speed by adding new weights connecting the context nodes to the output nodes.Using this approach to identify reheat steam temperature system,we can see that the Elman neural network can identify linear system as good as least square method.Because neural network have the abilities of self-learning and mapping arbitrary non-linear system,the simulation result of non-linear system identification use Elman neural network is nice too.更多还原