Predictive Control Algorithm Of Artificial Pancreas Interval Model Based On Multi-objective Optimization

At present,diabetes is treated by injecting insulin.Patients need to inject insulin multiple times a day,and the dosage of insulin cannot be precisely controlled,resulting in hypoglycemia.The artificial pancreas system is considered to be one of the most promising methods for the treatment of diabetes due to a completely closed-loop control system that automatically adjusts the rate of insulin infusion according to blood glucose levels.At present,the development level of the insulin pump and real-time continuous monitoring technology in the artificial pancreas system has met the clinical requirements.However,since the regulation of human blood glucose includes body fluid regulation and nerves,it is an extremely complex dynamic process,and the internal factors and blood glucose There are many external factors,so it is very challenging for the design of artificial pancreas closed-loop control algorithm.Aiming at the problems of no further optimization in the interval control of the existing model predictive control algorithm and no consideration of insulin economy,this topic proposes a diamond-shaped interval soft constraint multi-objective optimization model predictive control algorithm.At the same time,in the process of blood glucose regulation,both control performance and the impact of insulin cost must be considered.Therefore,for the construction and solution of the predictive control algorithm objective function,it is necessary to study the multi-objective optimization problem.The population multi-object fireworks algorithm solves the objective function.The specific research work is as follows:First,because the ARX model can approximately reflect the relationship between artificial pancreas blood glucose and insulin,the Auto Regressive with eXogenous Input(ARX)model is selected as the prediction model,and the ARX model and Kalman parameter estimation method are used to predict the ARX model.System identification,the identification data source is the artificial pancreas simulation platform certified by the US Food and Drug Administration(FDA)-simulation data generated by UVa / Padova at specific carbohydrate intake and insulin injection rate to determine the model order And model coefficients to provide a model basis for subsequent algorithm design.Secondly,in view of the problems that the existing model predictive control algorithm set value control leads to low degrees of freedom and interval control is not optimized within the interval,a diamond-shaped interval soft constraint model predictive control algorithm is established.The algorithm determines the diamond interval according to the set tolerance interval and the target value,so that the blood glucose value is preferentially controlled within the diamond interval,and then performs dynamic optimization,gradually controls the blood glucose value to the target value,and improves the control performance of the system.And on the UVa / Padova platform for experimental verification,through the average blood glucose and standard deviation and percentage of time evaluation indicators to verify the performance of the algorithm in controlling blood glucose levels.Finally,considering the two major issues of control performance and economic benefits in the process of blood glucose regulation,two objective functions are constructed in the predictive control algorithm,taking into account the effects of control performance and insulin cost.A multi-group multi-object fireworks algorithm(MP-MOFWA)is proposed to solve this multi-objective function.The algorithm constructs a multi-group,multi-crossing operator operation method to improve the diversity of the group;at the same time,according to the different contributions of the sub-population to the optimal solution set,the local and all search capabilities of the algorithm are improved.By comparing with other algorithms,the performance of the algorithm in terms of diversity and convergence is verified.The multi-objective optimized interval predictive control algorithm was experimentally verified on the UVa /Padova platform to verify the advantages of the algorithm in both control performance and economy.