基于改进粒子群算法的神经网络及其在核电站故障诊断中的应用

ANN based on an improved PSO algorithm and its application to fault diagnosis in NPP

针对传统BP神经网络以及标准PSO算法神经网络在故障诊断过程中易出现收敛速度低及易陷入局部最优点的问题,提出了一种使用改进PSO(IM-PSO)算法进行训练的神经网络。该方法具有收敛速度快、不易陷入局部极小值的优点。使用该方法对核电站二回路凝给水系统的典型故障进行了诊断,诊断结果表明:该方法的性能优于传统BP算法和标准PSO算法,在复杂系统故障诊断领域具有很好的应用前景。

As the traditional BP ANN and standard PSO-ANN tend to cause a low convergence rate and are prone to focus on the local optimization points in the course of fault diagnosis, a new ANN training method based on the improved PSO （IM-PSO） algorithm has been proposed. This method has advantages of fast convergence rate and not being prone to remain at local optimization points, so that the method is used for the fault diagnosis of the condensation and feed water system in nuclear power plants. The diagnosis results prove that this method ANN and standard PSO-ANN, thus it can be applied is in performance superior to the traditional BP to the fault diagnosis field of a complex system.