改进粒子群-极限学习机模型在面板堆石坝运行期沉降预测中的应用

Application of IPSO-ELM Model in Settlement Prediction of Face Rockfill Dam During Operation Period

针对极限学习机(ELM)沉降预测模型中随机权值和阈值导致部分节点无效的问题,引入改进粒子群算法(IPSO)优化极限学习机的参数,构建基于改进粒子群-极限学习机算法的面板堆石坝运行期沉降预测模型,并将其应用于某完建的面板堆石坝运行期沉降预测中。结果表明,与未优化的极限学习机预测模型和统计回归预测模型的拟合预测结果相比,经改进粒子群算法优化后的极限学习机预测模型在测点上的拟合精度更高,且由于引入改进粒子群算法后,极限学习机在满足精度条件下所需预设的隐含层神经元数更少,这可极大地降低模型网络的复杂度,避免模型在训练中出现过拟合现象;三个模型中IPSO-ELM模型的泛化能力更好,预测结果更精确、稳定。

Aiming at the problem that some nodes are invalid due to random weights and thre sholds in thesettlement prediction model of extreme learning machine(ELM),an improved particle swarm optimization(PSO)algorithm is introduced to optimize the parameters of the extreme learning machine(ELM),and a settlement prediction model of CFRD in operation period is established based on the improved PSO-ELM algorithm.And the model is applied to the settlement prediction of a completed CFRD during the operation period.Compared with the extreme learning machine prediction model and statistical regression prediction model,the prediction model optimized by the improved PSO has higher fitting accuracy.Due to the introduction of the improved PSO,the number of pre-set hidden layer neurons of the extreme learning machine is less,which greatly reduces the complexity of the model network and avoids the phenomenon of over fitting in the training of the model.The IPSO-ELM model has the best generalization capability and more stable prediction results.