农田有效灌溉面积的预测方法及应用  

Prediction Methods of Effective Irrigated Area and Its Application

作　　者：李建伟[1] 魏伟[2] 陈沛然[3] 袁志华[1] 

机构地区：[1]河南农业大学机电工程学院,郑州450002 [2]安徽工业大学机械工程学院,安徽马鞍山243002 [3]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《湖北农业科学》2013年第9期2157-2160,共4页Hubei Agricultural Sciences

摘　　要：以归一化处理后的1986-2010年河南省农田有效灌溉面积的统计数据作为样本数据,分别采用BP神经网络和支持向量机回归两种方法建立了农田有效灌溉面积的预测模型。预测结果表明,支持向量机的预测方法具有更高的预测精度和更强的泛化能力,预测误差仅为BP神经网络预测误差的11.8%,更适合进行农田有效灌溉面积的预测。最后采用两种模型分别对河南省"十二五"期间的农田有效灌溉面积进行了预测,指出了其变化趋势。Normalized statistical data of effective irrigated area in Henan province from 1986 to 2010 were used as samples. Prediction models based on BP neural network and support vector machine respectively was established. Predicted results showed that the method based on support vector machine had higher forecast precision and better generalization ability, fore- cast error was 11.8% of BP neural network＇s forecast error, was more suitable for predicting of effective irrigated area. Final- ly, effective irrigated area of Henan province in Chinese ＂the 12th 5-year-plan＂ was predicted respectively using two mod- els, and the trend was pointed out.

关 键 词：农田有效灌溉面积 BP神经网络 支持向量机 预测 

分 类 号：S279.2[农业科学—农业水土工程] TP183[农业科学—农业工程]