【作者】 邹健；

【导师】 诸静；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2002， 博士

【摘要】 预测控制是面向实际工业过程发展起来的一类先进控制方法，一直深受控制界的关注，将智能控制理论与预测控制机理相结合，使预测控制向智能化方向发展，以满足复杂工业过程控制的需要，是当前预测控制发展的新趋势。本文针对工业控制过程中的实际问题对智能预测控制进行了深入的研究，提出了几种新的智能预测控制算法，并结合浙江省重点科技项目，以水泥生产过程中关键性工艺环节为对象，设计了一类新的智能预测控制器。实际运行结果表明，该算法能获得良好的控制效果。总结全文，论文的主要内容可概括如下： 1．在介绍了T-S模糊模型和GPC基本原理的基础上，将基于T-S模型的GPC归纳为三种算法，从理论上对这三种算法进行了详细地推导，并通过仿真研究比较了三种算法的控制性能和计算负担上的差异；从而为这一类模糊预测控制的实际应用提供了选择的依据，也为进一步的性能分析奠定了基础。 2．根据水泥生产线系统复杂、干扰频繁、全局模型建立困难的实际情况，通过对回转窑烧成工艺和分解炉过程特性的分析，提出一种基于T-S模型的多模结构的预测控制算法，并在DCS系统中开发了实时控制的软件。实际应用表明，对于存在着大时滞和参数时变的水泥回转窑系统，该算法不但能获得良好的控制精度(温度波动不超过850±10％℃)，而且能明显地提高水泥熟料的分解率。 3．灰色预测是根据被控对象的动态行为和运行趋势进行分析的预测方式。它具有建模数据量小，计算简单的特点，适用于不确定性复杂系统动态预测。在介绍了灰色系统建模理论的基础上，将灰色模型引入到预测模糊控制中。针对灰色预测步长对控制性能影响进行了详细分析，在此基础上，首次提出一种基于预测步长自调整的灰色预测模糊控制器。通过仿真研究验证了该方法能明显地改善常规模糊控制和固定预测步长的灰色预测模糊控制效果。 4．神经网络预测控制是智能预测控制的一个重要组成部分。通过对PH中和过程的分析，针对常规DMC方法在控制该过程中存在的问题，提出一种基于CMAC的非线性DMC控制算法，并通过仿真验证了该算法能有效地克服常规 浙门刁、学协l学位论义 仰I儿 D＊C方沿存在的不足，有效地改故了控十帅l能。。5．实！讨性和收敛性的矛盾是N祈￥遗传算法在非线性预测控制滚动优化叶。应 用的一大戒题，针对这一问题，本文对标准迅传算法进行了改进，首次提山一fZ。 基于启发式遗传算法。根扼被控对象的状态信；③、，在实时优化中适时调整控制显 的搜索区域，并将这种方法应用于基于＊＊＊C网络的非线性预测控制中。仿真 结果表明，该优化方法能较好地克服了标准遗传算法所存在的缺陷，不仅实时性 较强，控制精度也有显著提高，从而为提升遗传算法在非线性预测控制中的适用 性作出了有益的探索。更多还原

【Abstract】 Intelligent predictive control has started to present as one of the new research domain of model predictive control. The studies and applications of intelligent predictive is discussed in this dissertation. Several new control strategies are present in this paper. In this paper, a kind of intelligent predictive control algorithm is designed and applied in Hangzhou cement plant. In conclusion the main contents are as follows:1. Based on the introduction of the principles of Takagi-Sugeno(T-S) fuzzy model and generalized predictive control(GPC) algorithm, the fuzzy predictive control method combining GPC and T-S model is classified as three kinds of algorithms. The design method of these algorithms is presented in detail. A comparison of these FGPC strategies in control performance and complexity of computation is given by simulation.2. The cement rotary kiln is a complex process, which has nonlinear and time-varying characteristic and exist strong and frequent disturbance, so global model is hard to establish. Based on analysis of the characteristics of the decomposing furnace in Hangzhou cement rotary kiln plant, a new fuzzy predictive control method with multi-model structure is proposed. The parameters of T-S fuzzy model can be modified, depending on the flow of raw material. Real time control software of this algorithm has been developed on the platform of Plantscape distributed control system of Honeywell Corporation. Application results exhibit the superiority of the proposed method over conventional ones even on the condition of existing large delay and time-varying parameters. It has improved the decomposing rate of cement from 81.5% to 89.1%, ensured the temperature varying within ?0% 癈.3.Grey prediction is a kind of effective prediction method, which only needs few output sample data (four data are enough) for modeling a system. Therefore, it is easy to model a system without complex computations. Based on the introduction of grey system modeling theory, a kind of fuzzy predictive controller using grey model isinproposed. A fuzzy decision making mechanism of the grey model ’s prediction step is applied to improve the performance of fuzzy controller. Simulation results show the superiority of the proposed method over the conventional fuzzy controller and the grey model predictive fuzzy controller with fix predictive step size.4. Neural net predictive control is an important branch of the intelligent predictive method. By considering the nonlinear characteristic of PH neutralization process, a nonlinear DMC control scheme, based on the CM AC model which is used to model the titration equation, is proposed to overcome the nonlinear disturbance caused by the unknown spices existing in the process stream. The effectiveness of this strategy has been verified via simulation.5. The conflict between convergent speed and solution accuracy has been the impediment when GA is applied in nonlinear predictive control. In order to overcome this problem, a heuristic generic algorithm (HGA) is proposed in nonlinear predictive controller based on CMAC predictive model. The main characteristics of HGA are that it can change search space online according to the process state while performing the online optimization. So it can enhance the convergent speed and solution accuracy simultaneously. Simulation results demonstrate the feasibility of the proposed algorithm.更多还原