兰州大学学报(自然科学版)
蘭州大學學報(自然科學版)
란주대학학보(자연과학판)
JOURNAL OF LANZHOU UNIVERSITY(NATURAL SCIENCES)
2013年
3期
337-339,346
,共4页
李世萍%孔令彬%肖玮%田梦%张文煜
李世萍%孔令彬%肖瑋%田夢%張文煜
리세평%공령빈%초위%전몽%장문욱
BP神经网络%风速观测资料%均一性%订正模型
BP神經網絡%風速觀測資料%均一性%訂正模型
BP신경망락%풍속관측자료%균일성%정정모형
BP neural network%speed observation data%homogeneity%correction model
利用甘肃省河西地区敦煌、酒泉、民勤3个站点2004?2007年自动观测与同期人工观测系统的风速观测资料,采用BP神经网络建立了人工观测风速观测资料序列的订正模型,并进行了模拟效果检验.结果表明:利用BP神经网络建立的订正模型能对风速观测资料进行较高精度的订正,3个站点风速拟合差值相对于原差值明显减小,订正结果与自动观测资料的相关系数均在0.90以上.各个站点的平均相对误差较小,在12%以下,且订正模型的稳定性和可扩展性较好;各个站点的平均相对均方根差为3.20~3.84,效果良好,可为建立均一性时间序列的风速观测资料提供参考.
利用甘肅省河西地區敦煌、酒泉、民勤3箇站點2004?2007年自動觀測與同期人工觀測繫統的風速觀測資料,採用BP神經網絡建立瞭人工觀測風速觀測資料序列的訂正模型,併進行瞭模擬效果檢驗.結果錶明:利用BP神經網絡建立的訂正模型能對風速觀測資料進行較高精度的訂正,3箇站點風速擬閤差值相對于原差值明顯減小,訂正結果與自動觀測資料的相關繫數均在0.90以上.各箇站點的平均相對誤差較小,在12%以下,且訂正模型的穩定性和可擴展性較好;各箇站點的平均相對均方根差為3.20~3.84,效果良好,可為建立均一性時間序列的風速觀測資料提供參攷.
이용감숙성하서지구돈황、주천、민근3개참점2004?2007년자동관측여동기인공관측계통적풍속관측자료,채용BP신경망락건립료인공관측풍속관측자료서렬적정정모형,병진행료모의효과검험.결과표명:이용BP신경망락건립적정정모형능대풍속관측자료진행교고정도적정정,3개참점풍속의합차치상대우원차치명현감소,정정결과여자동관측자료적상관계수균재0.90이상.각개참점적평균상대오차교소,재12%이하,차정정모형적은정성화가확전성교호;각개참점적평균상대균방근차위3.20~3.84,효과량호,가위건립균일성시간서렬적풍속관측자료제공삼고.
Using automatic and manual observation wind data from Dunhuang, Jiuquan and Minqin of Hexi area of Gansu Province from 2004 to 2007, a correction model for manually observed wind data sequence was established based on the BP neural network and the simulation effect was tested. The results show that the correction model based on the BP neural network gave a more accurate revision of the wind data. The fitting difference of wind speed was significantly reduced compared with the original difference. The correlation coe?cient of the revised results and the automatic observation data were above 0.90 and the average relative error was smaller, i.e. below 12 %. The stability and expandability of the correction model was better, the average relative root mean square error was 3.20~3.84 and the effect was good. This correction model can provide a reference for establishing the homogeneity of time series of wind data.