MEA-BP预测模型在水布垭面板堆石坝沉降变形预测中的应用  

Application of MEA-BP Forecast Model in Settlement Deformation Prediction of Shuibuya Concrete Face Rockfill Dam

作　　者：徐朗 蔡德所[1] 

机构地区：[1]三峡大学水利与环境学院,湖北宜昌443002

出　　处：《人民珠江》2018年第2期48-51,共4页Pearl River

基　　金：国家自然科学基金重点项目(NO.51439003)

摘　　要：由于单一的BP神经网络预测模型存在一些缺陷,对于复杂的非线性问题,BP神经网络预测模型的拟合能力具有局限性,容易陷入局部极小值,网络权值和阈值的选择具有随机性。运用具有全局寻优能力的思维进化算法(MEA)对BP神经网络的初始权值和阈值进行优化,建立MEA-BP预测模型。将单一的BP神经网络预测模型、基于遗传算法(GA)优化的BP神经网络预测模型与MEA-BP预测模型应用到实际工程中进行对比,预测结果表明MEA-BP预测模型在水布垭面板堆石坝沉降变形预测中的预测精度最高,效果最好,具有一定的实际应用价值。Because the single BP neural network prediction model has some shortcomings, for the complex nonlinear problems, the fitting ability of BP neural network prediction model is limited, which would easy to trap into local minimum, and the network weights and threshold selection would be randomness. The initial weights and thresholds of BP neural network would be optimized by using the mind evolutionary algorithm( MEA) with global optimization ability, and the MEA-BP prediction model was established. This paper had compared the actual effect of three models including the single BP neural network prediction model, the BP neural network prediction model with the Genetic Algorithm( GA) and the MEA-BP prediction model. The results showed that the precision of MEA-BP prediction model was the highest in predicting the settlement deformation of Shuibuya concrete faced rockfill dam. Also, it had the best effect and a certain practical application value.

关 键 词：大坝变形预测 坝体沉降 水布垭面板堆石坝 BP神经网络 思维进化算法 

分 类 号：TV641.43[水利工程—水利水电工程]