Gas Hydrate Equilibrium Prediction Research Based On Gray-RBF Algorithm

Gas hydrate formation process is a complex process affected by many factors, in which the most important influence is the gas composition, temperature and pressure. Because its generating process is random, weak dependence and mutation characteristics, besides many shortcomings of using of the previous thermodynamic phase equilibrium prediction, such as imprecise numerical prediction and slowly calculation speed. So it is of important practical significance to establish a mathematical model of prediction accuracy and little error using mathematical theory.Due to the above reasons, establishing the gas hydrate phase equilibrium prediction model by use of GM(1,N) model’s forecast the theory and RBF neural network with many advantages such as fast convergence speed and short learning time and high degree of approximation in this paper. GM(1,N) model with lower original data requirements can predict system controlled by multiple factors. The main research contents of are as follows.1.The gas hydrate phase equilibrium prediction model of Multiple factors is established, useing of GM (1, N) model of modeling multi variable systems and RBF neural network theory.2.The method of difference is selected in combining with the grey theory and RBF neural network. To improve the prediction accuracy of the model, the phase equilibrium is selected in the model calculation as well as parameters of practical value.3.RBF neural network prediction model of gas hydrate and grey-RBF neural network prediction model of gas hydrate and improved grey-RBF neural network prediction model of gas hydrate are established. The phase equilibrium of gas hydrate is predicted by the three models, improved grey-RBF neural network prediction model takes on best prediction accuracy and minimum error by the experimental analysis.4.The phase equilibrium parameters of gas hydrate formation of the single component gases of gas sample(CH4,C2H6 and C3H8) and multi component gas of gas sample (CH4,C2H6,C3H8,i-C4H10,n-C4H10,C02 and N2) are prediced by many sample data, and the forecast precision and relative error of the improved model have been analysised.The results show that estimates of the model meets the accuracy requirement. The phase equilibrium model using improved grey -RBF neural network will become a effective method of calculation and prediction for the phase equilibrium of gas hydrate.