人工神经网络及支持向量机在降雨量预报中的应用  被引量:11

Application of Artificial Neural Networks and Support Vector Machine in Rainfall forecating

在线阅读下载全文

作  者:张乐坚[1,2] 程明虎[2,3] 田付友[4] 

机构地区:[1]中国气象局中国气象科学研究院,北京100081 [2]南京信息工程大学,江苏南京210044 [3]中国气象局气象探测中心,北京100081 [4]中国气象局国家气象中心,北京100081

出  处:《高原气象》2010年第4期982-991,共10页Plateau Meteorology

基  金:国家自然科学基金项目(40575009);863计划(2007AA061901)共同资助

摘  要:使用误差反向传播网络(BPN)和约当网络(JN)两种人工神经网络(ANN)以及支持向量机(SVM)对降雨量进行了1h和3h预报的研究,并与交叉相关法(CCM)外推预报的结果进行了比较。针对安徽省2003年6~7月的降水过程,比较了网络(文中指BPN、JN和SVM)和CCM预报降雨量与实况降雨量的雨带分布、强降雨区域和强度;使用命中率(HR)、虚警率(FAR)、漏报率(NAP)、临界成功指数(CSI)、相关系数(CC)和均方根误差(RMSE)这6个指标并结合天气分析检验网络和CCM的预报效果。结果表明:网络和CCM对雨带和强降雨区域的预报比较准确,但是对强降雨中心位置和强度的预报与实况存在差异;在使用HR、FAR、NAP和CSI检验预报效果时设定的阈值对预报结果的评价有影响;预报的中小尺度结构与天气分析的结果一致;网络与CCM以及不同的网络之间的预报结果存在着差异;连续预报的结果表明,与CCM相比,网络对3h预报的效果优于1h的。The SVM (support vector machine) and ANNs (artificial neural networks) including-BPN (Back Propagation Network),JN (Jordan Network) are used to forecast-rainfall intensity fields in-a lead time of 1 and 3 hours comparing with CCM (Cross-Correlation Method).The evaluation is presented between the observation and the forecasting including rainfall zone,heavy rainfall region and intensity.The six indices containing-HR (Hit Rate),FAR (False Alarm Rate),NAP (No Alarm Probability),CSI (Critical Success Index),CC (Correlation Coefficient) and RMSE (Root Mean Square Error) are utilized-to evaluate the result-of forecasting.The performance of networks and CCM is compared with that of the weather analysis.The rainfall zone and heavy rainfall region of the forecasting are close to that of observation,while-there are some difference on-the exact position and intensity of heavy rainfall.Under the condition of HR,FAR,NAP and CSI-the threshold plays a key role on the evaluation and the smaller threshold-indicates better result.The results of forecasting are similar to that of the weather analysis.Furthermore,there is difference between networks and CCM.Comparing with CCM,the forecasting results of networks in-a lead time of 3 hours are better than those in a lead time of 1 hour.

关 键 词:人工神经网络 支持向量机 降雨量预报 

分 类 号:P457.6[天文地球—大气科学及气象学]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象