关键词: 核岭回归算法(KRR), 贝叶斯优化算法(BOA), 网格搜索法(GS), CO2溶解度, 溶解度图版, 碳捕集、利用与封存（CCUS）技术

Abstract:  In the process of CO2 injection, a threephase dynamic equilibrium of CO2, crude oil, and formation water is formed. The solubility of CO2 in oilwater system will directly affect the displacement effect and storage potential of CO2. In order to study the solubility of CO2 in oilwater system, by taking the oilwater system of Jilin oilfield as an example, the solubility experiment of CO2 in oilwater system with different volume proportion was carried out by using PVT analyzer, and the solubility rule of CO2 in oilwater system was clarified. In the experiment, the solubility of CO2 in oilwater system increased with the increase of saturation pressure and oilwater volume ratio. Based on the experimental data, the parameters of kernel ridge regression(KRR) were optimized by Grid search method (GS) and Bayesian optimization algorithm (BOA), respectively, and the CO2 solubility prediction model of CO2 in oilwater system was established. The average relative errors of GSKRR model and BOAKRR model in the optimization model based on KRR algorithm are 6.758% and 1.998% respectively. It shows that by this BOAKRR model , the solubility of CO2 in oilwater system can be predicted and plotted with higher precision. By using BOAKRR model, the solubility chart of CO2 in oilwater system under different temperature and oilwater ratio provides support for the application of CCUS（carbon capture, utilization and storage） technology.

Key words:  , kernel ridge regression(KRR), Bayesian optimization algorithm (BOA), grid search(GS) , CO2 solubility, solubility chart, carbon capture, utilization and storage technology