【作者】 薛亚丽；

【导师】 吕崇德；

【作者基本信息】 清华大学 ， 动力工程及工程热物理， 2005， 博士

【摘要】 热力过程本质上是复杂的多变量系统,输入变量和输出变量间复杂的耦合关系使得多变量控制系统的参数优化一直是影响其投运的关键和难点,因此研究通用有效的多变量控制系统的优化方法对提高热力过程运行的经济性、安全性具有重要意义。本文利用遗传算法研究了单变量系统和多变量系统的控制优化问题,主要内容包括:1.改进了单变量控制系统的优化及评价方法。将控制器参数稳定域和鲁棒稳定域的启发性信息与遗传算法相结合,提高了单变量控制系统遗传算法优化的搜索效率。采用Mento-Carlo方法对不确定系统的控制效果进行评价,为众多控制方法的选择建立了客观的评价标准。2.提出了一种基于完整性分析的多回路控制系统优化方法。利用遗传算法和完整性理论对多回路控制系统的参数进行同时优化,获得整体最优的控制效果。研究结果表明,优化后控制效果有明显改善,且优化方法只要求目标函数可数值计算,随着系统维数的增加,优化的工作量并没有显著增加,显示了良好的适用性。3.提出了一种多目标、多扰动、多模型的多变量控制系统参数优化方法。针对控制结构确定的多变量系统,适当定义包含多目标、多扰动、多模型的目标函数,形成优化问题并通过遗传算法求解。方法广泛适于解决复杂热力过程中具有综合控制要求和约束条件的控制系统参数设计问题。4.将上述方法应用到热力过程中的典型多变量控制系统——汽轮发电机组控制系统、机炉协调控制系统和气化炉标准控制问题中。针对各个系统的特点,分别采用适合工程实现的控制器结构,定义反映控制系统综合要求的目标函数,进行控制参数的优化。研究结果表明,本文的优化方法在取得优良控制效果的同时,兼具灵活性和通用性的特点,在多变量热力过程控制优化中有广阔的应用前景。最后,论文还定量地比较了理想继电反馈与饱和继电闭环反馈辨识方法对四类典型热工对象的辨识精度,获得了饱和继电反馈辨识提高精度的定量认识。更多还原

【Abstract】 Thermal power plants are essentially complicated multivariable systemsand have strong coupling among variables. It is clear that the parametersoptimization becomes the key difficulty in operation. So the study on a kindof general and effective optimization method for multivariable controlsystems has great significance for improving the economic and safe runninglevel of thermal power processes. Under the general framework provided bygenetic algorithm, this paper deals with the control system optimization ofsingle variable system and multivariable system. The main contributions ofthe paper are as follows:1. Improve the optimization and evaluation method of single variablecontrol system. By combining the information of stable parameteric regionand robust stable space with genetic algorithm, the search effiency of geneticalgorithm optimization on single variable control system is greatly improved.The robust performance of resulted uncertainty system is evaluated throughMento-Carlo method, and this evaluation criteria is comfortably extended toother control methods.2. Based on integrity analysis, a kind of optimization method formulti-loop control systems is proposed. To achieve the optimized performanceon the multi-loop control system, the genetic algorithm and integrity theoryare used simultaneously in parameter optimization. Simulation results showthat the performance of the optimized system has been improved notably. Itonly requires that the objective function can be calculated numerically, andwith the increase of system dimention, the workload of the optimization hasnot obviously increased.3. A kind of optimization method is proposed for multivariable controlsystem with multi-objective, multi-disturbance and multi-model. For thestructure-specified multivariable system, by defining the objective function,which contains the multi-objective, multi-disturbance and multi-model, theoptimization problem is formulated and solved using genetic algorithm. Theproposed method is generally suitable for parameter design of complicatedthermal process control system, which has hybrid control requirements andconstraints.4. The proposed method is applied to several typical multivariablethermal processes: turbine-generator sets control system, boiler-turbinecoordinated control system and gasifier benchmark problem. The controlstructure adopted is easy to realize in practice, and the objective functions aredefined to refect the integrated demands respectively according tocharacteristics of each system. The simulation results show that the proposedmethod is quite suitable for the optimization of multivariable thermalprocesses, owing to good control performance and the feature of feasibilityand generolity.And in the last part, a comparison between the identification accuracy ofsaturation relay and standard relay feedback idenfication method isquantitively drawn on four kind of typical thermal objects. Then it is clearwhen and how much the saturation relay feedback method can make animprovement on the identification accuracy.更多还原