神经网络组合模型在大坝位移预测中的应用

摘要

为了提高大坝位移量预测的精度,引入了粒子群优化神经网络的组合预测方法.该组合预测方法以灰色GM(1,1)、回归分析法的预测值及预测结果的平均值作为输入,实际的大坝位移量作为输出,进行非线性组合.实例表明,粒子群优化神经网络组合预测法的均方误差为1.194 6 mm,平均绝对误差为0.781 4 mm,均小于单一模型及等权平均模型的相应值,适用于大坝位移量的预测.%In order to improve the accuracy of dam displacement prediction, a combination forecasting method is applied of particle swarm optimization-neural network. In this method, the dam displacement predictions and the average forecast results of grey GM (1,1) and regression analysis are used as the network inputs, and the actual displacement value as the outputs to carry out nonlinear combination. The example shows that the mean square error, average absolute error are 1. 194 6 mm and 0. 781 4 mm of the particle swarm optimization-neural network combination forecasting method, less than the corresponding prediction error of a single model and the equal weight method, and is suitable for the dam displacement prediction.