改进灰狼群优化算法的环境污染物预测研究

Improved Grey Wolf Optimization Algorithm for Environmental Pollutant Prediction

针对传统RBF网络在环境污染物预测中出现的泛化能力弱和准确度低的问题,提出一种组合最近邻聚类算法(NNCA)和改进灰狼群(IGWO)的优化预测算法.首先,针对RBF网络中心参数学习不足,利用最近邻聚类算法(NNCA)调整RBF神经网络的聚类中心参数;其次,针对灰狼群算法寻优能力不足,利用sin函数对参数ɑ进行非线性调整,利用适应度加权系数进行位置调整,得到改进的灰狼群优化算法(IGWO),利用IGWO优化算法进行调整RBF神经网络的权值参数.最后利用NNCA-IGWO-RBF算法对草原环境中的PM10浓度进行预测,验证预测算法的有效性.结果表明,相对于传统的RBF和GWORBF算法,该算法预测误差最小,有更高的精确度和更好的泛化能力,能够为污染物治理提供指导作用.

Aiming at the weak generalization ability and lowaccuracy of traditional RBF network in environmental pollutant prediction,this paper proposes a combined nearest neighbor clustering algorithm(NNCA) and an optimized prediction algorithm for improved Grey Wolf Optimization(IGWO).Firstly,aiming at the insufficient learning of RBF network center parameters,the nearest neighbor clustering algorithm(NNCA) is used to adjust the clustering center parameters of RBF neural network.Secondly,aiming at the shortage of the search ability of Grey Wolf Optimization(GWO),using the sine function to nonlinear adjustment of parameters ɑ,and using fitness function weighting coefficient to position adjustment,improved Grey Wolf Optimization Algorithm(IGWO),using IGWO optimization algorithm to adjust the weight parameters of RBF neural network.Finally,the NNCA-IGWO-RBF algorithm is used to predict PM10 concentration in grassland environment,and the effectiveness of the prediction algorithm is verified.The results showthat compared with the traditional RBF and GWO-RBF algorithms,this algorithm has the smallest prediction error,higher accuracy and better generalization ability,and can provide guidance for pollutant control.