摘要
为了提高动态系统测量数据的准确性,提出了基于改进Elman网络的动态系统测量数据检验方法.采用一步前预测方法构造网络的训练样本,用带自适应学习率的动态BP算法进行网络的训练,通过训练后的网络对各测量参数进行估计,实现测量数据的在线检验.对某电厂CCS系统进行了仿真试验,结果表明该方法可避免形成简单的一一对应映射,能正确获取系统动态特性,具有较强的降噪能力,能够正确辨识出测量数据中存在的不良值,提高了系统监测的可靠性和健壮性.
A novel data validation method based on the modified Elman network is presented to detect gross errors in measured data of dynamic system. The structure of the dynamic recurrent network is given, the dynamic back propagation (BP) learning method is presented and the one-step-ahead method is adopted for creating training set. The simulation results of a coordinate control system (CCS) in power plant indicate that the dynamics of the system is achieved and the predicted value follows the real trend well. By the right detection of the invalid parts existing in the sampling data, the effectiveness and robustness of the whole system are improved.
出处
《东南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2005年第1期50-54,共5页
Journal of Southeast University:Natural Science Edition
关键词
ELMAN网络
动态系统
数据检验
电站
Backpropagation
Computer control systems
Data acquisition
Electric power plants
Learning algorithms
Measurement errors