基于混合神经网络的水合物无机盐抑制剂浓度预测

打开文本图片集

中图分类号：TE832 文献标志码：A

Abstract：Taking the predictionof NaCl concentration as an example，the temperature，pressure，and gas components processed by kernel principal component analysis（KPCA）were usedasinput parameters.The wavelet neural network（WNN） was utilized to predict the NaClconcentration，andthe WNNwas optimized bythe Genetic annealing algorithm（GASA）and AdaBoostalgorithm.Then，the AdaBoost-GASA-WNN prediction modelfortheconcentrationof the hydrate inhibitor NaCl was established. The results show that the man square error （ eMSE ） of the model is decreased by 3.87 after KPCA processing，and the eMSE of the optimized model is further reduced to 9.51. Compared with ELM，KNN，RF，and data fitting methods，the eMSE is declined by 5.8，17.74，2.91，and 8.81，respectively. And the prediction effect is the best.

Keywords：hydrate prevention；mineral salt inhibitor；neural network；optimization algorithm；AdaBoost algorithm

天然气水合物堵塞是海洋油气钻井过程中经常 遇到的问题，同时也是石油天然气管道运输中最严重的故障之一[1]，会造成管道和设备的堵塞，甚至产生环境和安全问题，因此需要采用适当的方法来防止水合物生成。（剩余12020字）