摘要
针对火电厂实时成本建模问题,该文提出一种基于RBF网的在线建模方法,即资源优化网络,简称RON。RON能自动根据最近一段时间内的误差信息优化网络结构:如果当前网络不能实现新输入样本,则在线增加新隐节点,否则使用梯度法在线调节网络的隐节点数据中心和扩展常数。数据中心和扩展常数调整过程中还引入了隐节点的合并操作和删除操作。进一步介绍了火电厂实时成本建模方法,包括网络结构的确定,及如何获取训练样本。仿真实例表明,RON对能较好地适应对象参数的时变特性,并对训练样本集变化具有较好的鲁棒性。
For realtime cost modeling in power plant, a Radial Basis Function Neural Network based online modeling method for nonlinear dynamics systems, called Resource-Optimizing Network (RON), is developed in this paper. According to newest error information, RON optimizes system resource automatically: When current network performs worse on new data sample, a new hidden unit is added, otherwise gradient method is used to adjust unit centers and spread constants. Hidden units combination and deletion in the process of adjusting data centers and spread constants are also introduced. Then, based on RON, a realtime cost modeling method is given, including introduction of the model stmcture and the acquisition of training samples. Simulation shows that RON can trace the time-varying of model parameters effectively, and performs better robustness against the changing of training sets.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2004年第7期246-252,共7页
Proceedings of the CSEE
关键词
火电厂
RBF网络
火电机组
实时成本
在线建模
Thermal power engineering
Realtime cost
Online modeling
Reural networks
RBF nets
Resourceoptimization network
Generalization ability
