摘要
泥石流的流速预测是泥石流灾害防治的核心问题之一。由于泥石流流速的影响因素众多,需要寻求能够综合反映泥石流流速影响因素的预测模型。移动最小二乘法(MLS)预测模型具有自学习和自组织及捕捉到影响因素数值微小变化的能力,可以解决泥石流流速预测存在的一些问题。采用云南蒋家沟泥石流流速实测数据作为训练样本和预测样本,以泥深、比降、密度、颗粒的平均粒径作为输入因子。讨论了用MLS方法进行泥石流平均流速预测的可行性与有效性,并将预测结果与经验公式、BP神经网络以及支持向量机进行了对比。结果表明,MLS方法的最大预测误差为4.6%,平均误差为2.7%,预测精度优于经验公式、BP神经网络及支持向量机方法。MLS方法可以为泥石流防治提供更准确的科学依据。
Forecast for velocity of debris flow is one of the core issues in prevention and control of debris flow. Because of the velocity affected by many factors, it is need to seek a prediction model which can reflect comprehen-sively factors influencing the average velocity. The prediction mode based on moving least square （MLS） method possesses the ability of self-leafing and self-organization. The model also has capacity of catching the subtle variation of the influencing factors. It can solve the existing problems in prediction. The model was established by training and forecasting the observation data of Jiangjia iGully debris flow in Dongchuan, Yunnan Province, taking the debris flow depth , the gradient of the channel, the debris flow density and the average grain size as the input factors. The feasibility and availability of prediction of debris flow arerage velocity by MLS method were discussed and the forecasted results were compared with the computed results from empiric equation, BP neural networts and SVM, Results show that maximum error of MLS prediction model is 4.6% ,the average error is 2.7% and the prediction accuracy of MLS prediction model is higher than those of empiric equation, BP meural metwork and SVM. The MLS prediction model can offer more accurate scientific basis for prevention and control of debris flow.
出处
《自然灾害学报》
CSCD
北大核心
2016年第4期56-61,共6页
Journal of Natural Disasters
基金
国家自然科学基金项目(41272336)~~
关键词
泥石流
蒋家沟
平均流速
MLS方法
预测
debris flow
Jiangjia Gully
average flow velocity
moving least square（MLS） method
forecast
