基于径向基神经网络的盐水中CO2溶解度模型

Prediction of Dissolution of CO2 in Brine Based on RBF Neural Network

确定盐水中CO2的溶解量对CO2地质封存潜力和外溢风险评估至关重要,经典溶解度模型建立在热力学平衡定律上,具有一定的限制性,而RBF径向基人工神经网络具有很强的泛化能力,能够关联复杂变量之间的映射关系.利用收集到的实验数据建立了RBF神经网络用于预测盐水中CO2的溶解度,和之前已经建立的BP神经网络模型对比,将实验数据、RBF神经网络模型、BP神经网络预测结果、PR-DUAN模型以及亨利定律计算值做了对比,为确定盐水中CO2的溶解度提供了一种新的RBF神经网络预测模型.

The dissolved amount of CO2 in brine is vital to the assessment of CO2 geological storage potential and spillover risk.The solubility of CO2 in different salinity and temperature and pressure conditions can be measured by experiments,which can also be predicted by using relevant solubility models.The classical solubility model based on the thermodynamic equilibrium law has some limitations,while RBF radial basis function artificial neural network can correlate the mapping relationship between complex variables.,and RBF model has a strong generalization ability.In this paper,RBF-ANN is established to predict the solubility of CO2 in brine based on the experimental data collected from literature.The RBF-ANN model is compared with the BP-ANN model established by the author before.The results of experimental data,RBF-ANN model and BP-ANN model are compared.In this work,a new RBF netrual network model is provided to predict the solubility of CO2.