RLS-BP人工神经网络算法在降水预报中的应用  

APPLICATION OF RLS-BP ARTIFICIAL NEURAL NETWORK ALGORITHM TO RAINFALL FORECAST

作　　者：余运河[1] 胡邦辉[2] 孔玉寿[2] 谭言科[2] 许云海 

机构地区：[1]总参气象中心 [2]解放军理工大学气象学院 [3]辽阳市65521部队测绘气象室

出　　处：《气象科学》2004年第3期333-341,共9页Journal of the Meteorological Sciences

基　　金：国家自然科学基金项目 (4 0 175 0 14 )资助

摘　　要：RLS -BP是一种新发展起来、自动设置网络主要参数和具有快速收敛能力的人工神经网络模型 ,用预报对象动态消空处理加以改进 ,应用于大连地区短中期降水MOS预报 ,实施了其预报效果与普通BP网络对比的试验。结果表明 ,改进后的新模型既提高了降水的预报精度 ,又有良好的计算稳定性 ,并且适合推广应用于气温、能见度等诸多气象要素的预报和预测问题。As one of the new revised artificial neural network models with setting the primary parameters automatically and the rapid convergence, RLS-BP is used in the short-mid range rainfall MOS forecast in Dalian after improving the dynamic elimination for predictands. A comparison of RLS-BP to the ordinary BP has been carried out,too. The results showed that the new network enhances not only the rainfall forecast accuracy, but also the calculation stability,and can be extended to the forecast/prediction of other meteorological elements, such as air temperature, visibility etc.

关 键 词：数值预报产品解释应用 RLS BP网络 动态消空指标库 降水 预报 

分 类 号：P458[天文地球—大气科学及气象学]