基于PSO算法的BP神经网络对水体叶绿素a的预测

BP神经网络(Back Propagation Network)在水体富营养化评价及预测中已广泛应用,但传统BP算法的收敛速度慢并易陷入局部最优.提出了一种基于微粒群(PSO)算法的BP神经网络模型,利用PSO对神经网络的权值进行修正,优化神经网络结构及算法全局收敛性.选择最能代表明湖水质状况的5号采样点作为研究对象,把2009年4月—2010年3月的月样本插值为周样本,对明湖ρ(Chla)的短期变化趋势进行了预测,并用6号采样点数据来验证网络的泛化能力.比较分析基于PSO算法的新模型与传统BP算法模型的预测精度表明,新模型有效克服了传统算法的缺点,提高了网络的预测能力和学习能力.

Prediction of Chlorophyll a by BP Neural Network based on PSO Algorithm

Back Propagation(BP) neural networks have been widely used in the evaluation and prediction of water eutrophication.However,the traditional BP algorithm with limited optimization of the algorithm system has slow convergence.In this investigation,a model of the BP neural network based on the Particle Swarm Optimization(PSO) algorithm was proposed.The structure of the neural network and the overall convergence of the algorithm were optimized according to the adjusted weight value of the neural network by utilizing the PSO algorithm.Taking the fifth sampling location data representing the quality conditions of Ming Lake as the research object,short-term variation trends of chlorophyll a mass concentration were predicted.Monthly samples collected from April 2009 to March 2010 from Ming Lake were interpolated as weekly samples,and the BP neural network generalizations were verified using data from the sixth sampling location.Results showed that,by comparing the prediction accuracy of the BP neural network based on the PSO algorithm with the traditional BP algorithm,the new model effectively overcame the shortcomings of the traditional algorithm and improved the predictive power and learning ability of the BP neural network.