Study On Theoretical Model Of Sulfur Solubility In Gas Reservoir With High Sulfur Content

During the development of high sulfur gas reservoirs,sulfur deposition will occur with the decrease of temperature and pressure.Sulfur deposition will cause porosity and permeability of the formation to decrease,thus affecting the productivity of gas wells.The solubility of elemental sulfur in high sulphur gas is the precondition and basis for the study of sulfur deposition mechanism,sulfur deposition prediction and treatment technology.It is very important for the safe and efficient development of high sulphur gas reservoirs.At present,the main methods to determine the solubility of elemental sulfur in high sulphur gas are as follows:experimental method,empirical formula method,thermodynamic model,artificial neural network model and so on.The experimental method has high cost and great danger,and the empirical formula is easy to calculate but its scope of application is relatively narrow.The thermodynamic model needs two yuan interaction coefficient between sulfur and each component,and the model precision is not ideal.Although artificial neural network can simulate complex nonlinear mapping,it is easy to appear optimal,poor generalization ability and network structure.It is difficult to choose the parameters.Therefore,based on the least squares support vector machine(LSSVM),the solubility model of elemental sulfur in pure H2S and high sulfur content natural gas is established by combining the genetic algorithm(GA)and gray wolf algorithm(GWO),and the Chrastil sulfur solubility model is improved by the method of density inflection point.Finally,the three models are compared.It provides technical guidance for the development of high sulfur gas reservoirs.The main understanding is as follows:(1)The solubility model(GA-LSSVM)of sulfur in pure H2S and high sulphur natural gas,respectively,was established by using the method of LSSVM and genetic algorithm(GA).Using the same literature experimental data,the results show that GA-LSSVM model can predict the solubility of sulfur in pure H2S and high sulfur natural gas in the range of temperature 303-433 K and pressure 7.03-60MPa,and the absolute relative deviation(AARD)is 4.31%and 5.4%,respectively.Finally,the new model is verified and evaluated.(2)The solubility model of sulfur in pure H2S and Gao Hanliu natural gas(GWO-LSSVM)was established by the combination of LSSVM and Grey Wolf algorithm(GWO).Using the data of 239 experimental solubility of sulfur in pure H2S and high sulphur natural gas,the GWO-LSSVM model was trained(168)and predicted(71).The results show that the GWO-LSSVM model can predict the solubility of sulfur in pure H2S and high sulfur natural gas in the range of temperature 303-433K and pressure 7.03-60MPa,and AARD is 3.89%,4.55%respectively.Finally,the new model is validated and evaluated,and the temperature and pressure range applicable to the model are given.(3)Using the idea of density inflection point,the Chrastil association model is improved by piecewise fitting,and one or several density inflection points can be found in any high density gas,and then the solubility data of sulfur in pure H2S and high sulfur content natural gas are correlated.The results show that the new model AARD is 9%and has a certain ability to predict the solubility of sulfur with 4 kinds of semi empirical association models published at home and abroad.(4)Using the same literature experiment data,we compare the new GA-LSSVM model,GWO-LSSVM model,improved semi empirical association model,Hu Jinghong et al model,Guo Xiao et al model,Bian Xiaoqiang et al model,and Bian Xiaoqiang et al model.The results show that the calculation precision of Bian Xiaoqiang et al.(2009)is the highest,AARD=3.86%,Hu Jinghong model The calculation precision is the lowest,AARD=15.6%.The order of precision from high to low is:Bian Xiaoqiang et al.(2009),GWO-LSSVM model,Guo Xiao et al.Model,GA-LSSVM model,Bian Xiaoqiang et al model(2011),improved association model,Hu Jinghong et al.