【作者】 刘志远； 吕剑虹； 陈来九；

【Author】 Liu Zhi-yuan, Lu Jian-hong,Chen Lai-jiu Department of Power Engineering, Southeast University, Nanjing 210096; Nanjing Institute of Technology, Nanjing 210013

【机构】 东南大学动力工程系；

【摘要】 基于M-RAN算法的RBF神经网络是一种动态神经网络,适合于过程的在线建模,本文对M-RAN 算法的删除策略进行了改进,不仅删除那些连续对网络输出贡献较小的隐层单元,同时还将相似的隐层单元合并,使网络结构更加紧凑。将基于这种算法的RBF神经网络用于电厂热工过程非性线模型的在线辨识, 仿真研究表明了这种建模方法的有效性,且所得模型精度高,计算量小,可直接应用于基于模型的控制算法。更多还原

【Abstract】 M-RAN algorithm based radial basis function (RBF) neural network is a dynamical neural network and suitable for process online modeling. The pruning strategy of M-RAN algorithm is improved in this paper, not only to delete the hidden neurons continuous contributing little to the network output, but also to combine the similar hidden neurons and thus to implement a more compact network structure. The online identification of power plant thermal process nonlinear model is carried out with the method, simulation study indicates the validity of the method and established model with higher precision , less calculation and can be used in model based control algorithm directly.更多还原