基于在线辨识的特征模型预测函数控制研究

Research of Characteristic Model Predictive Functional Control Based on On - Line Identification

工业生产过程中运用模型控制问题,对于传统的预测函数控制,一旦预测模型与实际生产过程失配就会造成控制系统的控制性能下降。由于传统的预测函数控制算法中的预测模型通常是离线获得的,因此模型易出现失配。即使采用在线辨识方法来获取预测模型,由于在线辨识需花费时间,影响系统的实时性。通过构建实时数据库的监控层,实时采集生产过程控制参数,在监控层利用在线辨识方法辨识出被控过程的特征模型,然后将辨识出的特征模型作为预测函数控制算法的预测模型传至过程控制回路。采用获得在线辨识预测模型,不会影响系统的实时性。实验结果表明,根据在线辨识的特征模型与实际被控过程的失配很小,并提高了系统的控制性能。

Once predictive models do not match the practical process,the control system can not acquire better performance for the traditional predictive functional control strategy.Usually predictive models are acquired off-line,it is certain that predictive models do not match the practical process.Even if on-line identification is used to obtain predictive models,the timeliness of system may suffer some impact because of the identification time.By buliding the system monitor layer based on real-time database and acquiring the process parameters real-time,the characteristic model of process can be obtained using the identification method in the system monitor layer.The characteristic model,as the predictive model of predictive functional control,downlinks the control system.Using the predictive model from the on-line identification of the system monitor layer can not affect the system timeliness.The experimental results prove that using the new method,the predictive models can match the practical process and acquire better performance.