Research On An Adaptive Predictive Control Algorithm And Its Simulation In Intelligent Well Control

In actual industrial process, problems like nonlinear, parameters time-varying, or large interference, which bring challenges for the control algorithm design, are widespread, and the current solution for these types of problems is adaptive control algorithms. To improve the control performance, adaptive control combined with predictive control which has good industrial usability, and neural network technology which has better approximation of nonlinear functions is researched in this paper.Generalized predictive control can adjust the model parameters online adaptively, but the structure of model cannot be changed. The model output error cannot be offset no matter how parameter changes, when the structure of model exits dismatch with the structure of plant. This paper presents an adaptive predictive control algorithm, which combine the generalized predictive control and neural networks. The algorithm uses generalized predictive control to update model parameters, and uses neural network model to identify output error, then uses output error to calibrate model output. The whole process is self-adaptive. Simulation results show that the algorithm has good control effect.Then the adaptive predictive control algorithm proposed in this paper is applied to intelligent well control systems to control the down-hole flow rates of each segment. Because of the lack of condition, a numerical model of the reservoir and intelligent well systems are used to instead of practical fields. Eclipse software is selected to simulate the actual production process, and experiment results show algorithm proposed in this paper has the feasibility and application potentiality in intelligent well control system.