时滞约束系统的神经动态优化模型预测控制

Model predictive control of time-delayed restraint system based on neurodynamical optimization

将时滞约束系统的模型预测控制优化问题描述为一个带约束的二次规划问题进行处理,并采用一个对偶神经网络进行在线求解。该神经动态优化方法充分发挥了神经网络并行、分布式处理的优点,优化速度快,能够用来求解各种复杂的带约束优化问题。实验研究表明,该方法具有较高的优化精度和优化速度,提高了模型预测控制的在线优化能力,能够扩展模型预测控制的应用领域。

In this paper, online optimization problem of model predictive control （MPC） of time-delayed system with restraints is described as an quadratic programming （QP） problem with restraints and a dual neural network is used to solve this problem. This neurodynamical optimization method exerts the advanteages of neural network that neural network can solve problems in parallelly and distributedly, and has fast optimization speed;and this method can be used to solve all kinds of complicated optimization problems with restraints. Experiment study results show that the proposed MPC method has good optimization precision and optimization speed, and this neurodynamical optimization method improves the online optimization capability of MPC. The proposed MPC method based on neurodynamical op- timization extends the application fields of MPC.