摘要
采用基于实数编码的加速遗传算法(RAGA)代替传统的最小二乘法以优化GM(1,1)模型参数,并与BP人工神经网络相组合,形成了基于RAGA的等维灰色递补BP神经网络预测模型。运用此模型对三江平原创业农场地下水埋深进行动态预测,BP神经网络结构确定为3∶12∶3,预测结果的相对误差只有2.33%,与传统的GM(1,1)模型和BP神经网络模型预测结果相比,预测精度显著提高。通过此模型预测,从2007年到2012年,该地区地下水平均年下降0.3 m。
Replacing Least Square Method by Real coding based Accelerating Genetic Algorithm, the parameters of time response function in the GM （ 1,1 ） Model are optimized. Combined with BP Artificial Neural Networks Model, the Equal-dimension Gray Filling BP Neural Networks Model Based on RAGA is established. By this model, predicted the groundwater depth of Chuangye Farm in the Sanjiang Plain. The structural of BP Neural Networks is 3： 12： 3. The relative error is only 2.33%. Comparing with the traditional GM（ 1,1 ） Model or BP Neural Networks Model, the precision is highly increased. The result shows that the groundwater deep will descend 0. 3m in average annually in the area.
出处
《地理科学》
CSCD
北大核心
2009年第2期283-287,共5页
Scientia Geographica Sinica
基金
国家自然科学基金(No.30400275)
黑龙江省攻关项目(黑龙江省青年科学基金,No.QC04C28)资助
