基于SAPSO的灰色神经网络优化城市林研究  

Research on optimization of urban forest based on grey neural network of SAPSO

作　　者：罗美淑[1] 刘世勇[2] 孙强[1] 夏春艳[1] Luo Meishu;Liu Shiyong;Sun Qiang;Xia Chunyan(Mudanjiang Normal University,Mudanjiang,157012,China;Heilongjiang Preschool Education College,Mudanjiang,157011,China)

机构地区：[1]牡丹江师范学院,黑龙江牡丹江157012 [2]黑龙江幼儿师范高等专科学校,黑龙江牡丹江157011

出　　处：《中国农机化学报》2018年第8期76-80,共5页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金(61573362);黑龙江省自然科学基金(F2016039);黑龙江省青年科学基金项目(QC2013C067);牡丹江市科学技术计划项目(Z2016s0027);牡丹江师范学院科研项目(YB2018004)

摘　　要：为构建合理的城市生态系统,亟待预测适宜的城市林(城市地带性植被)。城市林的预测是一个复杂的非线性问题,其发展有波动性,选择合理的拟合方法可以提高预测精度。以东北地区的城市林为例进行研究,筛选7个影响城市植被类型的因子,以传统的灰色神经网络模型为基础,用粒子群算法初始网络参数,用模拟退火代替粒子群进行梯度修正,建立基于模拟退火算法(SA)和粒子群算法(PSO)的灰色神经网络模型。实验结果表明,改进后的模型预测拟合精度较高,残差均值为0.13,为城市林的预测提供一条新途径。In order to construct a reasonable urban ecosystem,it is urgent to predict suitable urban forests（urban zonal vegetation）.The prediction of urban forest is a complex nonlinear problem.Its development is fluctuating.Choosing a reasonable fitting method can improve the prediction accuracy.In this paper,we take the urban forest of Northeast China as an example to select seven factors that affect the type of urban vegetation.The grey neural network model based on simulated annealing algorithm（SA）and particle swarm optimization（PSO）is established on the basis of the traditional grey neural network model,using the initial network parameters of particle swarm optimization and simulated annealing instead of particle swarm optimization.The experimental results show that the improved model has a higher prediction precision and the residual mean value is 0.13,which is better than the existing model,and provides a new way for urban forest prediction.

关 键 词：城市林的预测 灰色系统 粒子群算法 模拟退火算法 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]