基于Volterra模型的预测控制及应用

Volterra model-based model predictive control and its application

由于工业实践的需要,非线性预测控制近年来受到广泛地关注.Volterra模型是一类特殊的非线性模型,非常适合描述工业过程中的无记忆非线性对象.传统的基于Volterra模型的控制器合成法及迭代计算预测控制器法计算量大,且不便于处理控制约束.非线性模型预测控制求解是典型的非线性规划问题,序列二次规划(sequential quadratic program,SQP)算法是求解非线性规划问题常用方法之一.针对Volterra非线性模型预测控制求解问题,本文将滤子法与一种信赖域SQP算法相结合,提出一种改进SQP算法用于基于非线性Volterra模型的带控制约束的多步预测控制求解,并分析了所提方法的收敛性.工业实例仿真结果证实了所提方法的可行性与有效性.

Because of the need of industrial application, nonlinear model predictive control has been concerned widely. Volterra models are a class of special nonlinear dynamic models and are suitable for describing the memoryless nonlinear dynamic process. Volterra model-based nonlinear predictive controller design is a typical nonlinear programming problem. But the traditional iterative computation method for solving the Volterra model-based nonlinear predictive controller needs a large amount of calculation to deal with the constraints. Sequential-quadratic-programming （SQP） is one of the adequate classical methods for solving the nonlinear programming problem. Combining filter technique and the trust region SQP, we propose an improved SQP algorithm to solve the constrained nonlinear predictive control problem based on Volterra model. The convergence property of the proposed method is also proved. The proposed method has been applied to control the melt index of the polyethylene process. The application results validate the feasibility and effectiveness of the proposed method.