基于模型移植的人工胰脏闭环控制研究
发布时间：2018-10-13 07:18
【摘要】：糖尿病已经逐渐发展成为一个危害人类健康和社会的全球性疾病,人工胰脏系统被认为是治疗糖尿病最有前景的方法之一。随着连续血糖监测系统和胰岛素泵的发展,人工胰脏系统的硬件部分已经满足临床要求,但是由于人体血糖调节过程是一个非线性、强扰动和非稳态的动态过程,再加上高低血糖风险的不对称性,给人工胰脏系统中的闭环控制算法设计带来了巨大的挑战。在传统的血糖控制算法领域中,在建模方法上,传统的建模方法需要利用对象的充足数据进行模型辨识,这样在采集数据时,会给患者和医生带来较大的负担,整体建模代价较高;在闭环控制算法上,传统的控制算法更多的是考虑算法上的创新,而没有过多的考虑高低血糖的风险不对称性。因此,如何利用先验知识,降低个性化建模的开销以及如何减少患者低血糖事件,是本文工作的重点和出发点。基于上述出发点,本文主要进行了以下研究:(1)在本课题前人工作的基础上,考虑血糖控制模型相关约束等因素,提出了一种面向血糖控制的PSO模型移植建模方法。该方法通过利用先验基模型知识,只需少量激励数据即可完成对象的个性化建模,克服了传统建模方法中需要等待充足激励数据才能建立个性化模型的问题。(2)提出了基于PSO模型移植建模方法的zone-MPC血糖闭环控制算法,实验结果证明,该方法仅仅使用少量激励数据(两小时)进行建模,就可实现与使用大量激励数据(五天)的ARX建模方法相同的控制性能。这种基于基模型的建模算法具有较高的建模效率和经济性,特别是在缺少患者血糖数据时,本文提出的PSO模型移植算法可以有效的替代传统的个性化建模方法,再结合zone-MPC控制算法,可以实现基本的全自动血糖闭环控制。(3)考虑到高低血糖的风险不对称性,提出了 一种基于不对称风险函数和PSO模型移植算法的zone-MPC血糖闭环控制算法,通过设计不对称风险函数来调节MPC目标函数中的高低血糖和控制量的权重系数,最后通过实验证实了该算法的有效性,特别是在大饮食场景下,降低了患者出现低血糖的频率。
[Abstract]:Diabetes has gradually developed into a global disease that endangers human health and society. Artificial pancreas system is regarded as one of the most promising methods for the treatment of diabetes. With the development of continuous blood glucose monitoring system and insulin pump, the hardware part of artificial pancreas system has met the clinical requirements, but the regulation process of human blood sugar is a nonlinear, strongly disturbed and unsteady dynamic process. Combined with the asymmetry of high hypoglycemia risk, the design of closed loop control algorithm in artificial pancreas system is a great challenge. In the field of traditional blood sugar control algorithm, the traditional modeling method needs to use sufficient data of the object to identify the model, which will bring great burden to patients and doctors when collecting data. The cost of global modeling is high; in the closed-loop control algorithm, the traditional control algorithm is more to consider the innovation of the algorithm, but not too much to consider the risk asymmetry of high hypoglycemia. Therefore, how to use prior knowledge, reduce the cost of personalized modeling and how to reduce patients with hypoglycemia events, is the focus and starting point of this work. Based on the above starting point, this paper mainly carried out the following research: (1) on the basis of the previous work of this paper, considering the related constraints of blood glucose control model, a modeling method for blood glucose control oriented PSO model transplantation is proposed. By using the knowledge of priori basis model, the method can complete the personalized modeling of objects with only a small amount of excitation data. It overcomes the problem that the traditional modeling method has to wait for sufficient excitation data to set up a personalized model. (2) the closed-loop control algorithm of zone-MPC blood sugar based on PSO model transplantation modeling method is proposed, and the experimental results show that, Using only a small amount of excitation data (two hours), the method can achieve the same control performance as the ARX modeling method which uses a large amount of excitation data (five days). This modeling algorithm based on basic model has high modeling efficiency and economy, especially in the absence of patient blood sugar data, the PSO model transplantation algorithm proposed in this paper can effectively replace the traditional personalized modeling method. Combined with zone-MPC control algorithm, the basic automatic blood glucose closed-loop control can be realized. (3) considering the risk asymmetry of high and low blood sugar, a closed loop control algorithm of zone-MPC blood glucose based on asymmetric risk function and PSO model transplantation algorithm is proposed. The asymmetric risk function is designed to adjust the weight coefficients of high hypoglycemia and control quantity in MPC objective function. Finally, the effectiveness of the algorithm is verified by experiments, especially in the large diet scenario, which reduces the frequency of hypoglycemia in patients.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318.1;R587.1
本文编号：2267757
[Abstract]:Diabetes has gradually developed into a global disease that endangers human health and society. Artificial pancreas system is regarded as one of the most promising methods for the treatment of diabetes. With the development of continuous blood glucose monitoring system and insulin pump, the hardware part of artificial pancreas system has met the clinical requirements, but the regulation process of human blood sugar is a nonlinear, strongly disturbed and unsteady dynamic process. Combined with the asymmetry of high hypoglycemia risk, the design of closed loop control algorithm in artificial pancreas system is a great challenge. In the field of traditional blood sugar control algorithm, the traditional modeling method needs to use sufficient data of the object to identify the model, which will bring great burden to patients and doctors when collecting data. The cost of global modeling is high; in the closed-loop control algorithm, the traditional control algorithm is more to consider the innovation of the algorithm, but not too much to consider the risk asymmetry of high hypoglycemia. Therefore, how to use prior knowledge, reduce the cost of personalized modeling and how to reduce patients with hypoglycemia events, is the focus and starting point of this work. Based on the above starting point, this paper mainly carried out the following research: (1) on the basis of the previous work of this paper, considering the related constraints of blood glucose control model, a modeling method for blood glucose control oriented PSO model transplantation is proposed. By using the knowledge of priori basis model, the method can complete the personalized modeling of objects with only a small amount of excitation data. It overcomes the problem that the traditional modeling method has to wait for sufficient excitation data to set up a personalized model. (2) the closed-loop control algorithm of zone-MPC blood sugar based on PSO model transplantation modeling method is proposed, and the experimental results show that, Using only a small amount of excitation data (two hours), the method can achieve the same control performance as the ARX modeling method which uses a large amount of excitation data (five days). This modeling algorithm based on basic model has high modeling efficiency and economy, especially in the absence of patient blood sugar data, the PSO model transplantation algorithm proposed in this paper can effectively replace the traditional personalized modeling method. Combined with zone-MPC control algorithm, the basic automatic blood glucose closed-loop control can be realized. (3) considering the risk asymmetry of high and low blood sugar, a closed loop control algorithm of zone-MPC blood glucose based on asymmetric risk function and PSO model transplantation algorithm is proposed. The asymmetric risk function is designed to adjust the weight coefficients of high hypoglycemia and control quantity in MPC objective function. Finally, the effectiveness of the algorithm is verified by experiments, especially in the large diet scenario, which reduces the frequency of hypoglycemia in patients.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318.1;R587.1
【参考文献】
相关期刊论文 前3条
1 杨志寨;;糖尿病治疗现状[J];中外医疗;2011年27期
2 常向云;范玲;孙侃;;糖尿病的治疗现状与糖尿病教育[J];医学与哲学(临床决策论坛版);2006年03期
3 高桂革;;模糊控制理论及其应用的发展[J];上海电机学院学报;2005年05期
，本文编号：2267757
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2267757.html
