CAJ | 학술논문

BP（Back Propagation）网络在用于水质预测时，存在运算速度慢和易陷入局部最优的缺点，与传统的BP网络相比，广义回归神经网络(General Regression Neural Network，GRNN)的计算速度快，预测精度较高。光滑因子σ是广义回归神经网络的唯一待确定参数,它对网络的预测性能影响很大,本文采用量子粒子群算法（quantum particle swarm algorithm，QP-SO）优化算法对光滑因子进行估算,并通过GRNN构建水质预测模型。实验表明：该模型能较好地预测氨氮变化趋势，为科学管理水质提供必要依据。
BP（Back Propagation）망락재용우수질예측시，존재운산속도만화역함입국부최우적결점，여전통적BP망락상비，엄의회귀신경망락(General Regression Neural Network，GRNN)적계산속도쾌，예측정도교고。광활인자σ시엄의회귀신경망락적유일대학정삼수,타대망락적예측성능영향흔대,본문채용양자입자군산법（quantum particle swarm algorithm，QP-SO）우화산법대광활인자진행고산,병통과GRNN구건수질예측모형。실험표명：해모형능교호지예측안담변화추세，위과학관리수질제공필요의거。
In the forecast of water quality, the BP network computes slowly and is tend to be trapped by local optima. Compared with BP network,the Generalized Regression Neural Network (GRNN) technology has many merits,such as computing quick-ly and a higher accuracy. Smoothing factorσ, the sole undetermined parameter of GRNN, has great influence on the prediction performance of the network. In this paper, quantum particle swarm algorithm (QPSO) is used to estimate the smoothing factor. Then build prediction model by GRNN. Numerical experiment results show that the new model can predict the trend of ammo-nia content better than the BP network do, and provide the necessary basis for scientific management of water quality.