设为首页
加入收藏
一种改进的BP算法及在降水预报中的应用
An Improved BP Algorithm and Its Application to Precipitation Forecast
-
摘要: 传统BP（back propagation）算法在实际应用中具有网络结构参数和学习训练参数难以确定、泛化能力差、训练学习易陷入局部极小点等问题。该文在传统BP算法的基础上，提出一种改进算法，在训练过程中能自动确定各种参数，并避免陷入局部极小点，提高网络的泛化能力。利用2003—2005年5—9月中国国家气象中心T213的数值预报产品，通过动力诊断得出反映降水的物理量，然后从中挑选出与降水关系较好的25个因子，连同中国国家气象中心T213模式、日本气象厅业务模式和德国气象局业务模式相应的降水量预报结果作为预报因子。采用改进的BP算法建立江淮流域68个站24 h降水 (08:00—08:00，北京时）3个等级（降水量≥0.1 mm，降水量≥10 mm，降水量≥25 mm）的预报模型。通过对2006—2007年5—9月68个站试报结果表明：改进BP算法对降水预报的TS评分大大高于传统BP算法，也高于几种模式的降水预报结果，同时，改进算法使降水预报的平均空报率、漏报率明显降低。Abstract: Objective forecast of precipitation is difficult because of its complex nonlinear characteristics. In order to enhance the ability of forecasting precipitation, artificial neural network (ANN) method is applied in numerical weather products interpretation. Among different types of ANN, the back propagation (BP) neural network is the most popular and influential one. However, traditional BP algorithm has some limitations such as the difficulties in determining network structure and the learning parameters, poor generalization ability and possibility of misleading to local minimum in learning process, etc. To resolve these problems, an improved algorithm is proposed.Based on T213 numerical forecast products of National Meteorological Center from May to September during 2003—2005, 25 factors are selected in terms of dynamic diagnostic analysis and statistical methods. The precipitation forecasts of operational global models from China National Meteorological Center, Japan Meteorological Agency and German Meteorological Administration are studied. Using the reformative BP algorithm, three grades forecast (≥0.1 mm, ≥10.0 mm, ≥25.0 mm) models are built to forecast 24 hour precipitation of 68 stations over Jiang Huai Basin. During the training process, precipitation samples are randomly divided into two kinds according to a certain proportion, training samples and testing samples. They are used to train the network and to check the error of output respectively so that all parameters are confirmed. By repeating training and learning of network, an optimal network model is obtained. The optimized forecast model is used to forecast precipitation of different grades, times and stations, from May to September during 2006—2007. The forecasting results of improved BP algorithm are compared with those of tradition BP algorithm and numerical models outputs. The average threat score (TS) of improved BP algorithm is the highest; the average false alarm rate (FAR) and missing alarm rate (MAR) of improved BP algorithm are much lower than the others. So the improved BP algorithm is superior and it indicates a potential for more accurate precipitation forecasting.
-
图 1 BP训练过程中陷入局部极小点示意图
Fig. 1 A schematic figure for the immer sion of the web to the local least solution in the BP training process
图 2 网络模型的泛化能力示意图
Fig. 2 A schematic figure of gene ralization ability for a network model
图 3 2006-2007年5-9月各方法的降水量预报检验评分对比
(a) 平均TS评分，(b) 平均空报率，(c) 平均漏报率
Fig. 3 Comparison of forecast verification for 24-hour precipitation among severalf orecasting methods from May to September during 2006-2007
(a) the average threat score (TS), (b) the ave rage false alarm rate (FAR), (c) the average missing alarm rate (MAR)
图 4 2006-2007年5-9月各方法对不低于0.1 mm降水预报TS评分对比
Fig. 4 Comparison of threat score (TS) for 24-hour precipitation (≥0.1 mm) among severalforecasting methods from May to Septembe rduring 2006-2007
图 5 2006-2007年5-9月各方法对不低于10.0 mm降水预报TS评分对比
Fig. 5 Comparison of threat score (TS) for 24-hour precipitation (≥10.0 mm) among several for ecasting methods from May to September during 2006-2007
-
[1] Hall T, Brooks H E, Doswell C A.Precipitation forecasting using a neural network.Weather and Forecasting, 1999, 14 :338-345. doi: 10.1175/1520-0434(1999)014<0338:PFUANN>2.0.CO;2 [2] Jin L, Ju W M, Miao Q L.Study on ANN-based multi-step prediction model of short-term climatic variation.Advances in Atmospheric Sciences, 2000, 17 :157-164. doi: 10.1007/s00376-000-0051-4 [3] Marzban C, Witt A.A bayesian neural network for severe-hail size prediction.Weather and Forecasting, 2001, 16 :600-610. doi: 10.1175/1520-0434(2001)016<0600:ABNNFS>2.0.CO;2 [4] 施丹平.人工神经网络方法在降水量级中期预报中的应用.气象, 2001, 27(6):40-42. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX200106007.htm [5] 张韧, 蒋国荣, 余志豪, 等.利用神经网络计算方法建立太平洋副高活动的预报模型.应用气象学报, 2004, 11(4):474-483. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20000469&flag=1 [6] 李法然, 周之栩, 陈卫锋, 等.湖州市大雾天气的成因分析及预报研究.应用气象学报, 2005, 16(6):794-803. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=200506103&flag=1 [7] 赵声蓉.多模式温度集成预报.应用气象学报, 2006, 17(1):52-58. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20060109&flag=1 [8] 胡文东, 沈桐立, 丁建军, 等.卫星资料的非线性反演同化与一次强降水预报试验.应用气象学报, 2006, 25(2):249-258. http://www.cnki.com.cn/Article/CJFDTOTAL-GYQX200602010.htm [9] 刘德, 李晶, 刘永华, 等.BP神经网络在长期天气过程预报中的应用实验.气象科技, 2006, 34(3):250-253. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKJ200603002.htm [10] 赵声蓉, 裴海瑛.客观定量预报中降水的预处理.应用气象学报, 2007, 18(1):21-28. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20070104&flag=1 [11] 叶笃正, 曾庆存, 郭裕福.当代气候研究.北京:气候出版社, 1991 :164-177. [12] 刘还珠, 赵声蓉, 陆志善, 等.国家气象中心气象要素的客观预报———MOS系统.应用气象学报, 2004, 15(2):181-191. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20040223&flag=1 [13] 陆如华, 何于班.卡尔曼滤波方法在天气预报中的应用.气象, 1994, 20(9):41-46. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX409.008.htm [14] 寗春蓉, 冯汉中.利用卡尔曼滤波方法释用数值预报产品.四川气象, 2004, 24(2):16-19. http://www.cnki.com.cn/Article/CJFDTOTAL-SCCX200402004.htm [15] 陈豫英, 刘还珠, 陈楠, 等.基于聚类天气分型的KNN方法在风预报中的应用.应用气象学报, 2008, 19(5):564-572. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20080507&flag=1 [16] 曾晓青, 邵明轩, 王式功, 等.基于交叉验证技术的KNN方法在降水预报中的试验.应用气象学报, 2008, 19(4):471-477. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20080411&flag=1 [17] 陈永义, 俞小鼎, 高学浩, 等.处理非线性分类和回归问题的一种新方法 (Ⅰ)———支持向量机方法简介.应用气象学报, 2004, 15(3):345-354. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20040344&flag=1 [18] 冯汉中, 陈永义.处理非线性分类和回归问题的一种新方法 (Ⅱ)———支持向量机方法在天气预报中的应用.应用气象学报, 2004, 15(3):355-365. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20040345&flag=1 [19] 赵国令, 肖科丽.支持向量机在天气预报中的应用.陕西气象, 2004(6):1-4. http://www.cnki.com.cn/Article/CJFDTOTAL-SLXB200606006.htm [20] 燕东渭, 孙田文, 杨艳, 等.支持向量机数据描述在西北暴雨预报中的应用试验.应用气象学报, 2007, 18(5):676-681. http://www.cnki.com.cn/Article/CJFDTOTAL-YYQX200705013.htm [21] 罗坚.基于SVM的分类方法及其应用.浙江工贸职业技术学院学报, 2008, 8(3):29-33. http://www.cnki.com.cn/Article/CJFDTOTAL-ZJGM200803009.htm [22] 曹均阔, 舒远仲, 叶水生.一种结构自组织的改进BP网络.计算机工程, 2005, 31(17):165-167. http://www.cnki.com.cn/Article/CJFDTOTAL-JSJC200517061.htm [23] 王正武, 张瑞平, 刘松.BP网络的一种改进学习方法.数学理论与应用, 2005, 25(1):31-34. http://www.cnki.com.cn/Article/CJFDTOTAL-LLYY200501007.htm [24] 李宏刚, 吕辉, 李刚.一种BP神经网络的改进方法及其应用.中国工程科学, 2005, 7(5):63-65. http://www.cnki.com.cn/Article/CJFDTOTAL-GCKX200505010.htm [25] 雷景生, 康耀红.一种改进的BP算法.海南大学学报自然科学, 2003, 21(4):326-329. [26] Rumelhart D E, Hinton G E, Williams R J.Learning Inter-nal Representation by Error Propagation.Parallel Distributed Processing:Explorations in the Microstructure of Cognition, 1986 :318-362. https://www.researchgate.net/publication/230876435_Learning_Internal_Representation_by_Error_Propagation [27] Lippmann R P.An introduction to computing with neural nets.IEEE ASSP Magazine, 1987, 4 :4-22. [28] Hecht-Nielsen R.Theory of the back-propagation neural net-work.IEEE International Joint Conference on Neural Networks, 1989, 1 :593-605. http://industry.wanfangdata.com.cn/dl/Detail/ExternalResource?id=dianzixb201412010%5e14 [29] Huang S C, Huang Y F.Bounds on number of hidden neu-rons in multilayer perceptrons.IEEE Transactions on Neu-ral Networks, 1991, 2 :47-55. doi: 10.1109/72.80290 [30] Reed R, Pruning algorithm-a survey.IEEE Transactions on Neural Networks, 1993, 4 :740-747. doi: 10.1109/72.248452 -
下载:
图(6)
计量
- 摘要浏览量: 4532
- HTML全文浏览量: 885
- PDF下载量: 2042
- 被引次数: 0
