基于BP神经网络连栽桉树人工林生长量预测  

Prediction for the Growth of Eucalyptus Plantations with Continuous-Planting Rotations Based on BP Neural Network

作　　者：龙滔[1] 覃连欢[1] 叶绍明[1] 

机构地区：[1]广西大学,南宁530004

出　　处：《东北林业大学学报》2012年第5期122-125,共4页Journal of Northeast Forestry University

基　　金：国家自然科学基金项目(31070560);广西自然科学基金项目(桂科字0991035);广西大学科研基金项目(X081003)

摘　　要：以广西国有东门林场雷卡分场的3个连栽代次的尾巨桉(Eucalyptus urophylla×E.grandis)人工林为研究对象,以林分的林龄和林分密度作为输入变量,分别以林分的平均胸径和树高为输出变量,构建了6个2∶n∶1的BP人工神经网络模型。用林分前5 a的数据对网络进行训练,第6、7年数据进行测试,经过大量训练选取最优模型后,得出以2∶2∶1的结构训练的模型最优,林分平均胸径的3个BP网络模型平均预测精度分别为99.09%、98.35%和96.37%,平均树高的3个BP网络模型平均精度分别为96.22%、96.48%和96.6%。回归分析证明模型的拟合效果良好。模型可用来分析、模拟和预测相似条件下桉树人工林林分随林龄增长整个生长阶段的生长量变化情况。Taking Eucalyptus urophyllaxE, grandis plantations in Dongmen Forestry Center of Guangxi Province as the research object, six 2 ： n ： 1 BP artificial neural network models were constructed using stand age and stand density of the planta- tions as input variables, and average diameter at breast height （DBH） and average tree height as output variables respec- tively. The BP network was trained with data from the first five years, and the network model was examined with data from the sixth and seventh years. The optimum model was selected after extensive training. Experimental results showed that the optimum network structure was 2 ： 2 ： 1. The average forecast precisions of average DBH for three models were 99.09%, 98.35%, and 96.37%, and the average forecast precisions of average tree height were 96.22%, 96.48%, and 96.6%, respectively. Regression analysis indicated that the six constructed models exhibited good fit to the data. The models can be applied to the analysis, simulation and prediction of the growth of E. urophyllaxE, grandis plantations of different ages under similar conditions.

关 键 词：BP神经网络 尾巨桉 人工林 生长量预测 

分 类 号：S792.39[农业科学—林木遗传育种]