【作者】 寇得民；

【导师】 李晓斌；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2008， 硕士

【摘要】 实际工业生产中普遍存在着多变量耦合现象,为了实现其精确控制,首先需要解决的是系统模型的准确性问题,相比传统的系统辨识方法,本文采用了粒子群优化算法的智能辨识方法,以辨识的系统模型为预测模型,提出了多变量系统的智能预测函数解耦控制方法。在多变量模型的辨识方面,本文应用了粒子群优化算法,定性地分析系统参数空间范围,把系统辨识问题转化为参数空间寻优,利用粒子群优化算法在寻优过程中有效的避免局部最优,在整个参数空间内并行寻找获得系统参数的最优解。通过对多种模型的仿真实验研究表明,粒子群优化算法在多变量系统模型辨识中优于遗传算法。在多变量系统的解耦控制方面,研究了多变量系统预测函数解耦控制算法。将多变量系统的解耦控制问题化简为若干个单变量系统的预测函数控制,采取分散优化策略代替整体优化,利用预测函数控制算法的特点,引入基函数增加了设计的自由度,减少了在线计算量,使参数设计和算法过程求解大为简化,然后再利用该算法获得一个解析的解耦控制量计算方程;其控制器参数均可离线计算,而且算法简单,便于实现复杂、高维多变量系统的解耦控制。通过对多变量一阶滞后耦合系统和二阶滞后耦合系统的仿真表明,该方法具有很好的解耦控制效果。在预测函数控制的模型失配问题方面,针对多变量系统中的随机性和不确定性造成的预测模型失配,提出了智能预测函数解耦控制方法,主要包括:模糊预测函数解耦控制、遗传算法预测函数解耦控制和粒子群优化算法预测函数解耦控制。模糊预测函数解耦控制方法就是将模糊数学的方法引入到预测函数解耦控制中,利用模糊建模的方法解决预测函数解耦控制方法中由于模型失配严重而造成的控制无法稳定等问题。遗传算法和粒子群优化算法在参数优化中具有很好的辨识优化效果,利用遗传算法和粒子群优化算法对失配模型进行优化处理,以优化的模型作为预测模型进行解耦控制,形成遗传算法预测函数解耦控制和粒子群优化算法预测函数解耦控制。这些方法能够有效地克服控制模型的失配及系统的不确定性,具有很好的实时性。通过对实际工业对象的仿真及与自整定PID控制的比较表明,该方法具有很好的控制效果。更多还原

【Abstract】 Multivariable coupling phenomenon is existed widely in industry practice. In order to control multivariable systems accurately, it is needed to solve accuracy problem of system models. The method of system identification for particle swarm optimization (PSO) algorithm is used by contrast with traditional system identification. Using identified system models as predictive models, the multivariable decoupling control method of intelligent predictive function is introduced.The PSO algorithm is used in multivariable model identification, which turned system identification problems into optimization problems in parameter space by qualitatively analyze the scope of system parameter space. The PSO algorithm can effectively avoid getting into local optimum and is used to obtain the optimal solution by searching in the whole parameter space in parallel. The simulations were done for different model examples. The experiment results know that PSO algorithm is an effective method that performed better than genetic algorithm (GA) for system identification.In the aspect of MIMO systems decoupling control, a multivariable predictive function decoupling control (PFDC) algorithm is proposed. This algorithm can decompose the decoupling control problem of an MIMO system into predictive function controls (PFC) of several SISO systems. The decentralized optimization method was adopted to deal with coupled variables instead of the whole optimization. Utilizing the characteristics of PFC, base functions increase freedom of design, reduce the online calculation amount significantly, and thus efficiently simplify parameter design and the calculation load. Then, an analytical linear decoupling control equation can be derived with the PFDC algorithm, and the controller parameters can be calculated off-line. Therefore, the obtained algorithm is simple and can solve complicated high dimensions multivariable coupling control problems. The simulations were done for multivariable coupling systems of the first order plus time delay and the second order plus time delay, which show that the proposed control system is efficient and effective.Regarding to the model mismatch caused by randomicity and uncertainty of multivariable system processes, the intelligent PFDC is proposed, which mainly include fuzzy PFDC (F-PFDC), GA-PFDC and PSO-PFDC.The method of F-PFDC is importing fuzzy mathematics into PFDC. By using the fuzzy modeling method, it solves the instability in control process which is caused by severe model mismatch. GA and PSO have excellent effect in parameter optimization, which can be applied to mismatched model optimization. Taking optimized models as predictive models, GA-PFDC and PSO-PFDC are composed. The mentioned methods effectively get ride of the model mismatch and system uncertainty, and have good real-time performance. Via the comparison of practical industry model simulation and self-turning PID control, the proposed control system is efficient and effective.更多还原