一类参数不确定时滞系统的智能控制应用研究  

Application study on intelligent control of a class of time delay systems with parameter uncertainty

作　　者：张堃[1] 费敏锐[1] 吴建国[2] 张培建[2] 胡志江[2] 

机构地区：[1]上海大学机电工程与自动化学院上海市电站自动化技术重点实验室,上海200072 [2]南通大学电气工程学院,南通226007

出　　处：《仪器仪表学报》2014年第6期1394-1401,共8页Chinese Journal of Scientific Instrument

基　　金：国家重大仪器开发专项(2012YQ15008703);南通市应用基础项目(BK2012030)资助

摘　　要：针对参数不确定时滞系统容易出现预测模型失效和不确定信息无法精确描述的问题,提出一种快速循环坐标下降支持向量回归(FCCDSVR)预测模型。以环模制粒机作为实验对象,利用循环坐标下降法(CCD)求解支持向量回归模型的目标函数,快速优化制粒机温度预测模型。利用环模制粒机喂料量和温度的耦合关系,提出并设计独立双通道输入模糊专家温度控制器,根据喂料量近似得到蒸汽调节量,再根据预测模型得到的温度偏差值进行精确蒸汽调节,从而有效克服对象参数不确定和时滞等问题。仿真实验表明,与现有的支持向量回归模型相比,FCCDSVR模型在大样本情况下,收敛速度更快,建模时间更短;将此模型融入的模糊专家控制方法比PID控制算法和模糊控制算法有更好的曲线跟踪效果,在实际DCS平台中温度误差可控制在±1.3%以内,较±5%的工艺容许偏差有较大提升。Aiming at the problems that in the time-delay systems with parameter uncertainty, the phenomena that the prediction models fail and the uncertain information can not be described accurately often appear. In this paper, we propose a fuzzy expert control method for the control of singular systems based on the prediction model of fast cyclic coordinate descent support vector regression （FCCDSVR）. Using a hoop standard granulator as the test example, we developed a cyclic coordinate decent method to solve the objective function of the support vector regression model and to optimize the temperature prediction model in a large scale of samples. Using the coupling relationship between the take-in amount and temperature of the hoop standard granulator, we designed a fuzzy expert control based temperature system with independent double-channel inputs. This system can be used to regulate the steam amount based on the take-in amount and to adjust the steam amount according to the temperature difference obtained from the prediction model, so that the problems of parameter uncertainty, time delay and strong coupling can be solved. The simu- lation experiments indicate that compared with classical support vector regression SVR model, the FCCDSVR has faster convergence speed and shorter modeling time under large sample condition; and compared with PID control algorithm and fuzzy control algorithm, the fuzzy expert control method fused with the proposed model has better curve tracking effect. In the DCS experiment platform, the temperature error can be controlled within ±1.3% , which is a large improvement compared with the specified temperature tolerance of ±5%.

关 键 词：时滞 参数不确定 循环坐标下降 支持向量回归 模糊专家控制 环模制粒机 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] TH86[自动化与计算机技术—控制科学与工程]