Economic Model Predictive Control Of Bi-hormonal Artificial Pancreas System

Diabetes threatens health of the patient seriously,and bring about huge financial burden to the patient.Blood glucose management is a daily challenge for type 1 diabetes mellitus.The artificial pancreas(AP)system,especially the bi-hormonal type that leads in insulin and glucagon synchronously,is regarded as a hopeful way to achieve automated regulation of blood glucose(BG).At present,the existing design of the blood glucose closed-loop control algorithm only focuses on the control performance,but ignore the importance of reducing the economic cost of the control process and improving the computational efficiency of the algorithm.The proposed design of this paper combines economic model predictive control,particle swarm optimization and event triggered control,and takes into account the control performance,the economic cost and the calculation efficiency together.First,the economic model predictive control algorithm based on particle swarm optimization(PSO-EMPC)is designed.Combined with the switching rules designed based on EMPC's economic cost function,the dual hormone control of blood glucose is achieved.Compared with the simulation of switching IMC-PID and switching MPC,switching PSO-EMPC has achieved a double improvement in control performance and economic performance.Compared with the switching IMC-PID,the total price of hormone decreased by 59.69%.Compared with the switching MPC,the total price of hormone decreased by 46.28%.Second,the PSO-EMPC and event triggered control are integrated to form the event-triggered economic model predictive control(ET-EMPC).The trigger condition is designed to control the calculation efficiency of the algorithm.The simulation results show that the enhanced algorithm greatly improves the calculation efficiency while guaranteeing the control performance and economic performance.The calculation efficiency increased by 51.14%compared with the pre-improved algorithm.In addition,robustness tests are performed for the two algorithms about several factors,including meal changes,hormone sensitivity changes,and individual variability.For different patients,the algorithm adjusts its changeable parameters according to the parameters of the patient's prediction model to enhance the adaptive ability of the algorithm.The simulation results show that the two algorithms designed in this paper have good robustness.