摘要
设计一种基于动态霍普菲尔德(Hopfield)人工神经网络(ANN)的核动力装置异常运行状态监测方法。通过ANN的在线训练,保证ANN模型能够始终跟踪核动力装置因运行工况变化而引起的动态特性变化,降低误诊断的概率。通过观察ANN预测输出值与实际装置输出值之间的加权平均方差,可以在较早时间内检测出参数异常变化的出现。以一回路压力为例,进行运行参数典型异常变化的检测仿真实验。结果表明,该方法在全工况范围内,具有良好的参数异常变化检测能力。
An abnormal operation status detection method based on dynamic Hopfield artificial neural network （ANN） is designed for nuclear power plants. By online training of the ANN, it can be ensured that the ANN can tail after the normal change of the dynamic characteristics of the NPP caused by the change of its operation state, so as to reduce the possibility of misdiagnosis. By observing the weighted mean square error of the ANN predictive output and the real output of the device, the abnormal change of the parameters can be detected in early time. Taking the primary loop pressure of a NPP as example, several tests are performed to validate the ability of the method to detect the operation parameter abnormal change. The results show that within the entire operation spectrum of the NPP, the method exhibits well faculty of the parameter abnormal change detection.
出处
《核动力工程》
EI
CAS
CSCD
北大核心
2013年第6期156-160,共5页
Nuclear Power Engineering
