摘要
将一种动态递归网络——Elman神经网络应用到凝汽器真空预测。通过实例计算,表明该方法能够较准确地预测凝汽器真空,并具有训练速度快、结构简单、精度高的特点,是一种行之有效的预测方法。同时,对反向传播(BP)神经网络算法会出现局部极小值,提出了利用粒子群优化算法的全局寻优能力优化Elman神经网络连接权值系数的方法。仿真结果表明,利用粒子群优化算法的Elman神经网络可以建立精度更高的凝汽器真空预测模型。
A dynamical recurrent neural network, i. e, Elman neural network, has been used to predict vacuum in the condenser, through calculation in real example,it shows that the said method can more accurately predict vacuum in the condenser, and boasting features of high training speed, simple struc- ture,and high accuracy, being an effective and feasible prediction method. At the same time,owing to occurrence of partial minimum values in back propagation （BP） algorithm,a method for optimizing the connecting weight value coefficient of Elman neural network has been put forward by using the global optimization capability of particle swarm optimization （PSO） algorithm. Results of emulation show that a model for more accurately predicting vacuum in the condenser can be established by using the PSO algorithm and Elman neural network.
出处
《热力发电》
CAS
北大核心
2010年第4期7-11,35,共6页
Thermal Power Generation
