【摘要】：針對太湖流域化肥用量和糧食產(chǎn)量數(shù)據(jù),利用BP神經(jīng)網(wǎng)絡算法,建立了糧食產(chǎn)量與化肥用量之間的關系模型,以指導化肥減施增效。共收集了1980—2014年共35 a太湖流域16個縣市每個縣市的單位面積化肥用量和單位面積糧食產(chǎn)量數(shù)據(jù)。通過自回歸滑動平均模型(ARMA),對兩類數(shù)據(jù)進行時間序列分析,對數(shù)據(jù)中存在的缺項進行了填補。實驗表明,對于單位面積糧食產(chǎn)量數(shù)據(jù),用ARMA(2,6)模型能夠達到較佳的填補效果,均方誤差小于0.2,R~20.85。對于單位面積化肥用量數(shù)據(jù),用ARMA(3,7)模型較優(yōu),均方誤差小于0.02,R~20.80。說明ARMA模型數(shù)據(jù)填補效果較好。將填補后的不同縣的數(shù)據(jù)通過BP神經(jīng)網(wǎng)絡建立模型,描述了各縣市單位面積化肥用量和糧食產(chǎn)量的關聯(lián)關系。實驗表明,該方法擬合的均方誤差小于0.12,R~20.80,說明BP神經(jīng)網(wǎng)絡是一種準確度較高的擬合方法。通過分析各縣擬合結果,表明化肥用量有閾值,化肥用量低于該閾值,糧食產(chǎn)量將會較快速增長,高于該閾值,糧食產(chǎn)量將不再增長,過多的施用化肥并不能取得高產(chǎn)。
[Abstract]:Based on the data of chemical fertilizer dosage and grain yield in Taihu Lake Basin, the relationship model between grain yield and chemical fertilizer dosage was established by using BP neural network algorithm to guide the reduction and efficiency of chemical fertilizer application. From 1980 to 2014, the data of chemical fertilizer per unit area and grain yield per unit area in 16 counties and cities of Taihu Lake Basin from 1980 to 2014 were collected. The time series analysis of the two kinds of data is carried out by using the autoregression moving average model (ARMA), and the missing items in the data are filled. The experimental results show that ARMA (2, 6) model can achieve better filling effect for grain yield data per unit area, the mean square error is less than 0.2, R 鈮，

本文編號：2522632