Simulation and Prediction of Injection Pressure in CO2 Geological Sequestration Based on Improved LSO Algorithm and BP Neural Networks

Abstract

In order to avoid the occurrence of caprock integrity damage and gas escape due to injection pressure overrun in CO2 sequestration, an optimized back propagation (BP) neural network model based on Monte Carlo simulation (MCS) and an improved lion swarm optimization (ILSO) algorithm is proposed for the maximum sustainable injection pressure prediction. Firstly, a hydromechanical model is constructed to simulate the damage changes of the reservoir caprock during injection by ABAQUS. Secondly, in view of the uncertainties of formation parameters that could lead to deviations between the model calculation results and actual geological condition, the MCS method is used, and then, the probability distribution interval of the maximum injection pressure with high probability of caprock failure under different formation parameters is obtained by MATLAB. This effectively reduces the uncertainty and improves the calculation accuracy. Finally, based on the numerical simulation results, the maximum injection pressure prediction model is constructed. Aiming at the problems of the sensitivity of the BP neural network to initial weights and its poor convergence, tent chaotic mapping and difference mechanism are introduced to improve the LSO algorithm. Following this, the neural network is optimized by ILSO algorithm whose superiority is verified through 8 benchmark functions, and the maximum injection pressure is effectively predicted according to porosity, permeability, and other parameters. Experimental results show that, compared to the other three optimization methods, the ILSO_BP model has a faster convergence speed, higher prediction accuracy, and stability, which can provide a powerful guide for the safe injection of CO2 and efficient sequestration management.