摘要
神经网络方法可用于动力学系统的参数估计，包括时域和频域建模。将单层神经网络用于单交量线性连续时不变系统的参数估计中，线性系统分别用状态方程及传递函数来描述。对于状态方程描述的线性系统，假设系统状态、状态的导数及系统输入已知，给出了相应的神经网络结构及学习算法用以估计系统参数；对于传递函数形式描述的线性系统，推导出了基于频域数据的单层神经网络结构，用以估计传递函数的参数。
A simple neural network structure and its learning algorithm are presented for the purpose of system identification of a linear time invariant system which is represented by the state space equation. By use of the conclusion on time domain, a single layer neural network structure can be deduced to estimate the parameters of a linear system based on spectral analysis with the frequency response data available. The application results using the neural network have the same precision as that obtained using the modelling algorithm in frequency domain. Some mathematical formulas are also included.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
1996年第4期1-6,共6页
Journal of Tsinghua University(Science and Technology)
关键词
神经网络
参数估计
状态方程
单层神经网络
neural networks
parameter estimation
state space equation
transfer function
