Improvement Of Particle Swarm Optimization And Its Application In Numerical Simulation Of Oil Reservoir

As a simple and effective optimization algorithm, particle swarm optimization algorithm(PSO) has become a research focus in optimization techniques since it was proposed. Since PSO’s advantages of simplicity, high accuracy, fast convergence and a global convergence capacity in solving single objective and multi-objective optimization problems, it has been promoted and applied widely. With the in-depth research and the promotion of application, some shortcomings have been shown off in PSO, such as the slow convergence speed in the late part of the algorithm, and falling into the local minima, and so on.An improved PSO has been proposed to solve the unconstrained optimization problems. The conjugate gradient method and the opposition-based learning strategy have been coupled into the iterative process of PSO, each particle has been classified according to its velocity. The active particles have been improved by the opposition-based learning strategy, which maintains the diversity of the group, and the sable particles have been corrected by conjugate gradient method to enhance the convergence rate. The numerical experiment results show that the improved PSO algorithm can converge to the optimal solution quickly, and has a high convergence accuracy. For the application of the proposed algorithm, a new modeling method has been given for dynamic system using the improved PSO. The injection rates and production rates in oil production process have been used to test the models. The experiment results show that the obtained models have a small dependency on data. The established models can simulate the dynamic relationship among injection rates and production rates accurately by using little historical data.A hybrid algorithm has been presented for solving the constrained optimization problems. In this algorithm, constrained optimization problem is transformed into a series of sub-bound constrained optimization sub-problems with the Augmented Lagrange multiplier method, and the sub-problems are solved by a PSO embedded conjugate gradient method, which has a high speed to search the optimal value. The numerical experiment results show that the hybrid algorithm has a fast convergence rate. In the application of hybrid algorithm, the modeling method has been promoted to MIMO. Modeling MIMO is converted into a constrained optimization problem which is solved by the proposed hybrid algorithm. Numerical simulation is performed using petroleum production data. And the experiment results show that the obtained model can simulate the dynamic relationship among injection rates and production rates accurately, and predict the production rates precisely.