论文部分内容阅读
本文提出一种基于多神经网络并行预测模型的多变量协调预测控制,利用预测误差实时反馈校正各个神经网络预测模型的参数,对于对象的时变、未建模干扰及模型失配引起的误差均有很好的适应性。针对存在耦合的被控对象,本文在优化性能指标中采用多变量协调优化策略,对被控变量集及操作变量集优化,使被控变量达到优化值和使部分操作变量达到优化值。在单步预测控制的基础上,提出基于多RBF神经网络并行预测模型的多变量协调预测控制,提高了预测控制的鲁棒性及抗干扰能力。将此方法应用于精馏塔控制中,在保证主要产品质量合格的前提下,对操作变量进行约束,使部分操作变量达到优化值,从而减少能耗,提高经济效益。仿真结果表明,基于神经网络预测模型的多变量协调预测控制具有很好的动态特性、鲁棒性及显著的节能降耗效果。
In this paper, a multi-variable coordinated predictive control based on parallel prediction model of multi-neural network is proposed. The parameters of each neural network prediction model are corrected by real-time feedback of predictive error. The errors caused by time-varying, unmodeled and model mismatch Have a good adaptability. In order to control the existence of coupled objects, this paper uses multivariable coordination optimization strategy to optimize the set of controlled variables and operating variables, so that the controlled variables reach the optimal values and some of the operating variables reach the optimal values. Based on the single-step predictive control, a multi-variable coordinated predictive control based on multi-RBF neural network parallel predictive model is proposed, which improves the robustness and anti-interference ability of predictive control. This method is applied to the control of the distillation column. Under the precondition of ensuring the quality of the main products, the operating variables are constrained so that some of the operating variables reach the optimized values, so as to reduce energy consumption and increase economic benefits. Simulation results show that multivariable coordinated predictive control based on neural network prediction model has good dynamic characteristics, robustness and significant energy saving effect.