摘要
通过8个高维复杂函数对一种新型仿生群体智能算法——狼群算法(WPA)进行仿真验证,并与粒子群优化(PSO)算法进行对比。针对BP神经网络易陷入局部极值及初始权阈值参数难以确定的不足,利用WPA算法优化BP神经网络初始参数,提出WPA-BP径流预测模型,以云南省龙潭站枯水期月径流预测为例进行实例验证,并与PSO-BP及BP模型进行比较。结果表明:1WPA算法收敛精度远远优于PSO算法,具有较好的计算鲁棒性和全局寻优能力;2WPA-BP模型预测精度优于PSO-BP及BP模型,具有较好的预测精度和泛化能力。利用WPA算法优化BP神经网络的初始权值和阈值,可有效提高BP神经网络的预测精度和泛化能力。
Through 8 high dimension complex functions of a new bionic swarm intelligence algorithms, the wolves algorithm( WPA) for simulation, and the particle swarm optimization( PSO) algorithm were compared. Aiming at the deficiency of BP neural network is easy to fall into local extremum and initial weight and threshold parameters were difficult to determine, using WPA algorithm to optimize BP neural network initial parameters, prediction model of WPA- BP runoff to Long Tan Railway Station in Yunnan Province dry period of runoff prediction was verified as an example, and compared with PSO- BP and BP model. Results shown that: the WPA algorithm convergence precision was much better than PSO algorithm, and had better computational robustness and global optimization ability. The WPA- BP model prediction accuracy was better than PSO- BP and BP model had better prediction precision and generalization ability. By using the WPA algorithm to optimize BP neural network's initial weights and threshold of BP neural network could effectively improve the prediction accuracy and generalization ability.
出处
《人民珠江》
2016年第1期55-59,共5页
Pearl River
关键词
径流预测
狼群算法
BP神经网络
参数优化
枯水期
runoff forecasting
Wolf pack algorithm
BP neural network
parameter optimization
dry season
