摘要
在热工过程模型辨识中,被控对象动态特性往往表现出非线性、慢时变、大迟延和不确定性等特点,这使得难以对其建立比较精确的模型。为了达到精确建模的目的,提出一种基于微分进化算法和径向基函数神经网络的辨识方法。该方法采用基于能量分布正交最小二乘学习算法的径向基函数(radial basis function,RBF)神经网络,通过改进的微分进化算法,对神经网络辨识系统进行参数优化,使RBF神经网络能够更快、更精确地逼近实际系统的输出,达到精确建模的目的。仿真结果表明,在采用改进的RBF网络对热工复杂对象进行辨识时,通过微分进化算法进一步确定其最佳参数,可以取得更好的辨识效果。
In thermal process model identification, the dynamic characteristics of controlled object often show non-linear, slowly time-varying, large delay and uncertainty, which makes it difficult to set up its more precise model. In order to achieve the purpose of accurate modeling, a new identification method with differential evolution algorithm and radial basis function （RBF） neural network was proposed. In this method, the RBF neural network based on energy distribution orthogonal least squares （EDOLS） learning algorithm was used. At the same time, an improved differential evolution （IDE） algorithm was used to optimize the parameters of neural network identification system so that RBF neural network can faster, more accurately approximate the actual target output. The simulation results show that when the improved RBF neural network is used to identify thermal complex objects, it will achieve better identification effectiveness if the IDE algorithm is used to determine the optimum parameters.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2010年第8期110-116,共7页
Proceedings of the CSEE
关键词
热工过程:系统辨识
微分进化算法
径向基函数神经网络
能量分布正交最小二乘算法
thermal process
system identification
differential evolution algorithm
radial basis function neural network
energy distribution orthogonal least squares algorithm
