ZHAO Peng,LIU Lin.Prediction of forest fire danger rating based on meteorological and space-time factors[J].Journal of Forestry Engineering,2018,3(03):102-110.[doi:10.13360/j.Issn.2096-1359.2018.03.017]





Prediction of forest fire danger rating based on meteorological and space-time factors
东北林业大学信息与计算机工程学院,哈尔滨 150040
College of Information and Computer Engineering, Northeast Forestry University, Harbin 150040, China
森林火灾等级 预测模型 气象因子 时空因子
forest fire class prediction model meteorological factors space-time factors
传统森林火险预测方法在特征选择上主要依据气象因子,未考虑时间因子、空间地理位置、地形地貌等因素,可能在林火预测中产生较大误差,传统预测模型在面对不同林区更大的数据集和预测任务时,无法快速自动调参和自适应调整输入特征维度。笔者引入时间因子和空间因子,丰富火险预测特征,以降低森林火险等级预测中产生的误差; 为解决传统火灾预测算法在大数据集和多分类问题上效率逐渐下降问题,提出核主成分分析算法(kernel principal component analysis,KPCA)和改进的极限学习机算法(extreme learning machine,ELM)相结合的森林火险等级预测模型。结果表明,该模型能有效提高森林火险等级预测的准确率和执行效率。相对传统预测模型,其准确率可达89%,在预测时间上也有一定优势。
The forest fire is a serious natural disaster, causing great damage to the economy and ecology. The meteorological factors affect the forest fire potential. Global warming may result in more forest fires, and the geographic location and time are also important factors for forest fire danger rating. Therefore, the empirical analysis of the relationship between forest fires and these factors is of great significance. Generally, the areas where forest fires occur have spatial heterogeneity. In order to make forest fire prediction free from geographical restriction, it is necessary to put forward a method to provide technical support for forest fire management in most areas. In this study, the space-time factor was introduced to the forest fire danger rating model based on the meteorological factors to enrich the predicted features and reduce the predicted error. The predicted algorithm was optimized to improve the model efficiency and accuracy. In this study, a new predicted model based on kernel principal component analysis(KPCA)and extreme learning machine(ELM)were used. First, we analyzed the deficiency by using principal component analysis(PCA)to deal with the general nonlinear problems. Then, we expounded the PCA method based on kernel function and applied it to reduce the dimension of forest fire danger rating index. Next, we used memetic algorithm to optimize the input weight and implicit layer deviation of the ELM, and classified the hidden layer nodes to reduce the training time of the ELM in order to obtain the best prediction model. Finally, we combined this model and cross-validation algorithm to rate forest fire danger. The experimental results showed that the model can effectively improve the accuracy of forest fire danger rating and has significant advantages in parameter selection and training speed.


[1] ZHONG M H, FAN W C, LIU T M, et al. Statistical analysis on current status of China forest fire safety[J]. Fire Safety Journal, 2003, 38(3): 257-269.
[2] 郭福涛, 苏漳文, 马祥庆. 大兴安岭塔河地区雷击火发生驱动因子综合分析[J]. 生态学报, 2015, 35(19): 6439-6488.
GUO F T, SU Z W, MA X Q. Comprehensive analysis of driving factors for lightning-caused fire in Tahe, Daxing'an Mountain[J]. Acta Ecologica Sinica, 2015, 35(19): 6439-6488.
[3] 杨久红, 王小增. 基于径向基网络和气象因子的森林火灾预测系统[J]. 消防科学与技术, 2010, 29(7): 602-605.
YANG J H, WANG X Z. Forest fire forecasting system based on radial basis network and meteorological factors[J]. Fire Science and Technology, 2010, 29(7): 602-605.
[4] 梁慧玲, 林玉蕊, 杨光. 基于气象因子的随机森林算法在塔河地区林火预测中的应用[J]. 林业科学, 2016, 52(1): 89-98.
LIANG H L, LIN Y R, YANG G. Application of random forest algorithm on the forest fire prediction in Tahe area based on meteorological factors[J]. Scientia Silvae Sinicae, 2016, 52(1): 89-98.
[5] 高学攀, 廖士中. 基于贝叶斯网络的林火概率预测系统设计与实现[J]. 计算机工程与应用, 2017, 53(13): 246-251.
GAO X P, LIAO S Z. Design and implementation of forest fire probability prediction system based on Bayesian network[J]. Computer Engineering and Application, 2017, 53(13): 246-251.
[6] JAISER S R, BARNES J D, BAKER S N, et al. A multiple regression model of normal central and peripheral motor conduction times[J]. Muscle and Nerve, 2015, 51(5): 706-712.
[7] 李栋, 张文宇. 基于ELM和FOA的股票价格预测[J]. 计算机工程与应用, 2014, 50(18): 14-18.
LI D, ZHANG W Y. Stock price forecasting based on ELM and FOA[J]. Computer Engineering and Applications, 2014, 50(18): 14-18.
[8] ZHANG W, JIANG Q, CHEN L, et al. Two-stage ELM for phishing web pages detection using hybrid features[J]. World Wide Web-internet and Web Information Systems, 2017, 20(4): 797-813.
[9] ERTAM F, AVCI E. A new approach for internet traffic classification: GA-WK-ELM[J]. Measurement, 2017, 95(1): 135-142.
[10] MATEUS P, FERNANDES P M. Forest fires in Portugal: dynamics, causes and policies[J]. Springer International Publishing, 2014, 19(2): 97-115.
[11] CORTEZ P, MORAIS A P. A data mining approach to predict forest fires using meteorological data[M]∥NEVES J(Ed.). Guimaraes, Portugal: New Trends in Artificial Intelligence, Proceedings of the 13th EPIA 2007-Portuguese Conference on Artificial Intelligence, 2007: 512-523.
[12] 许志卿, 苏喜友, 张颐. 基于支持向量机方法的森林火险预测研究[J]. 中国农学通报, 2012, 28(13): 126-131.
XU Z Q, SU X Y, ZHANG Y. Forest fire prediction method of support vector machine based on[J]. China Agricultural Science Bulletin, 2012, 28(13): 126-131.
[13] YOUSSEF S, ABDELAZIZ B. Prediction of forest fires using artificial neural networks[J]. Applied Mathematical Sciences, 2013, 18(8): 662-669.
[14] DIMA A. An interval tree approach to predict forest fires using meteorological data[J]. International Journal of Computer Applications, 2015, 132(4): 17-22.
[15] 张巍. 森林定位观测与森林火险预警建设-大兴安岭国家森林生态观测站为例[J]. 内蒙古农业大学学报, 2009, 30(1): 127-131.
ZHANG W. The relationship between forest fire danger forecast and local forest observation case on Daxing'anling Mountain of Inner Mongolia local observation station[J]. Journal of Inner Mongolia Agricultural University, 2009, 30(1): 127-131.
[16] 邸雪颖, 李永福, 孙建, 等. 黑龙江省大兴安岭地区塔河县森林火险天气指标动态[J]. 应用生态学报, 2011, 22(5): 1240-1246.
DI X Y, LI Y F, SUN J, et al. Dynamics of forest fire weather indices in Tahe County of Great Xing'an Mountains region, Heilongjiang Province[J]. Chinese Journal of Applied Ecology, 2011, 22(5): 1240-1246.
[17] 彭欢, 史明昌, 孙瑜, 等. 基于Logistic 的大兴安岭雷击火预测模型[J]. 东北林业大学学报, 2014, 42(7): 166-169.
PENG H, SHI M C, SUN Y, et al. Lightning fire forecasting model of Daxing'an Mountain based on logistic model[J]. Journal of Northeast Forestry University, 2014, 42(7): 166-169.
[18] 朱沛林, 史明昌, MIKE W, 等. 黑龙江大兴安岭雷击火概率预测模型研究[J]. 中南林业科技大学学报, 2014, 34(8): 82-85.
ZHU P L, SHI M C, MIKE W, et al. A preliminary study on lightning-caused fire probability prediction model for Daxing'anling forest region[J]. Journal of Central South University of Forestry and Technology, 2014, 34(8): 82-85.
[19] 郭福涛, 胡海清, 马志海, 等. 不同模型对拟合大兴安岭林火发生与气象因素关系的适用性[J]. 应用生态学报, 2010, 21(1): 159-164.
GUO F T, HU H Q, MA Z H, et al. Applicability of different models in simulating the relationships between forest fire occurrence and weather factors in Daxing'an Mountains[J]. Chinese Journal of Applied Ecology, 2010, 21(1): 159-164.
[20] AGEE J K, SKINNER C N. Basic principles of forest fuel reduction treatments[J]. Forest Ecology and Management, 2013, 211(1): 83-96.
[21] UCI Repository of Machine Learning Databases[DB]. Available from: ftp:∥ftp.ics.uci.edu/pub/machinelearning-databases.
[22] YANG C J, HUANG K X, CHENG H. Haptic identification by ELM-controlled uncertain manipulator[J]. IEEE Transactions on Systems, Man, and Cybernetics Society, 2017, 99(3): 1-12.


 YANG Yuan,LIU Wenjin*.Investigation of the relationship between office-chair Kansei image and modeling elements[J].Journal of Forestry Engineering,2016,1(03):139.[doi:10.13360/j.issn.2096-1359.2016.03.025]


收稿日期:2017-10-25 修回日期:2018-01-24
基金项目:国家林业局公益性行业专项(201504307); 教育部中央高校基本科研业务费专项基金(2572017EB09)。
作者简介:赵鹏,男,教授,研究方向为林业信息化、模式识别。E-mail: nefuzhaopeng@nefu.edu.cn
更新日期/Last Update: 2018-05-15