Research On Optimization Method And Applications For Steady State Model Of Production Process

The real time optimization (RTO) for production process is the frontal subject in the field of international process control. RTO searches the optimal set point by optimization method and set the system simultaneously to make it running optimally. The key to RTO is to optimize the steady state model to obtain the optimal set point. The efficient optimization methods for steady state model of production process and set point optimization for oil production process are mainly studied in this dissertation. The main achievements are as follows:A predictor corrector infeasible interior point method, which overcomes the shortcomings of active set strategies used in solving the sub-problems of sequential quadratic programming, is proposed for quadratic programming. The method only requires nonnegative initial point instead of feasible or even strictly feasible initial point. It is proved to be equivalent to the level-1 perturbed composite Newton method, which follows its fast convergence. The numerical experiments also show that the proposed method is stable and efficient.An infeasible interior point sequential quadratic programming method is proposed for general nonlinear programming problems, based on the above predictor corrector infeasible interior point method. The automatic switch mechanism between line search and quadratic search is given to obtain the step length. The selection rules for merit function and the damp BFGS update formula for Hessian matrix are discussed. The proposed method is also used to optimize the parameters of PID controller and the numerical experiments on optimal control and large scale optimization problems are done to show its efficiency.Coupling the local searching technique and evolutionary algorithm gives a novel hybrid method to solve the general nonlinear programming problems globally. The modified augmented Lagrange functions are formulated via the object function, equality constraints and inequality constraints. It follows a global search for the modified augmented Lagrange function by the particle swarm optimization method under the bound constraints. The Lagrange multipliers and penalty parameters are modified according to the globe information obtained by particle swarm optimization method. The comparison experiments show its superiority of the novel hybrid method to existing evolutionary algorithms.A steady state model of oil production process is established using the rich historical data accumulated during the oil production, whose paremters are obtained by the proposed infeasible interior point sequential quadratic programming method. Based on the proposed steady state model, a set point optimization problem is formulated by maximizing profits or minimizing the production cost. The above proposed novel hybrid method is hired to solve the set point optimization problem globally. The numerical simulation is also made to verify the proposed model and method, using a set of data obtained by a commercial software Eclipse on a heterogeneous reservoir Synfield.