利用BP神经网络进行水库滑坡变形预测

滑坡变形监测与预测是滑坡预警预报中一种非常重要的途径。文章首先简单介绍了神经网络的基本原理和学习算法，然后利用某水库滑坡24期的GPS地表位移监测数据及其诱发因素即水库水位、降雨等资料，采用BP神经网络模型对该水库滑坡变形进行建模，最后将6期水库水位、降雨等资料输入模型进行滑坡变形预测，结果表明预测结果与实测数据符合性好，总体上能较好反映变形趋势。

Reservoir landslide deformation forecast using BP neural network

Landslides deformation monitoring and forecast is a very important approach to early landslide warning and forecast. The basic principle and algorithm of BP neural network are briefly introduced in this paper. With 24 sets of GPS displacement monitoring data and the corresponding cause information, i.e. reservoir level and rainfall of a reservoir landslides, the landslide deformation model is constructed using BP neural network. The late 6 sets of cause information are put into the deformation model and the forecasted deformation is obtained. The results show that the forecasted deformation has a good fit performance with the fatual surveying deformation and can reflect the overall deformation trend. The results have referrence value for landslide deformation forecast.