摘要
针对BP算法存在的缺点,提出了在共轭梯度方向修正权值、使用概率接受原则确定目标函数值变化取舍的CP-BP算法。实例证明:CP-BP算法在不同网络模型和初值下,均有快速和全局收敛性能。将此网络运用于变压器故障诊断,能有效地改善DGA分析方法的适用性和判断性能。
Back Propagation as a popular supervised training algorithm, converges slowly and immerses in vibration frequently. This paper presents a CP BP algorithm, which corrects weights in conjugate gradient and permits the moderate increase of optimum function by probability accepting.The feasibility of the method is examined using examples.
出处
《中国电机工程学报》
EI
CSCD
北大核心
1999年第3期41-45,共5页
Proceedings of the CSEE
关键词
变压器油
溶解气体诊断法
CP-BP算法
试验
Artificial neural network Back Propagation algorithm Conjugate gradient Dissolved gas in oil analysis