| Type 2 diabetes accounts for more than 90% of the total number of diabetic patients.Its various complications bring profound physical and psychological distress to patients,also puts a huge burden on caregivers and the health care system.The emergence if artificial pancreas provides a broad prospect for the treatment of type 2 diabetes,and the blood glucose control algorithm applied to artificial pancreas has become a related research hotspot.In this paper,the mechanism of the human blood glucose regulation system and the pathological characteristics of type 2 diabetes are studied.A glucose-insulin physiological model of patients with type 2 diabetes is established.A model free adaptive predictive control algorithm with disturbance compensation and real-time parameter estimation for hysteresis,non-linearity,time variability and meal disturbance of the plant.The main contents are organized as follows:1)A glucose-insulin physiological model for type 2 diabetes patients is established.After the mechanism of the human blood glucose regulation system and the pathological characteristics of type 2 diabetes are analyzed,the model is established based on the Dalla Man type 1 diabetes physiological model.The improvements include: in insulin subsystem,a slow balancing tissue chamber is added,an accurate pancreatic insulin secretion model with insulin biphasic secretion and an incretin effect model are established,and insulin related parameters are adjusted;then a glucose-insulin model conforming to the pathological characteristics of type 2 diabetes is established.The validity of the model was demonstrated by fitting the data from two clinical trials.2)An extended state observer based discrete model-free controller is designed.Based on ultra-local modeling,a discrete extended state observer is designed,and the unmodeled part and interference in ultra-local model are estimated by the observer.Finally a discrete extended state observer based intelligent proportional integral differential controller(ESOiPID)is obtained.The simulations verify the good observation performance of the discrete extended state observer and promising control performance of ESO-iPID under large sampling intervals.3)A model free adaptive predictive control algorithm with disturbance compensation is designed.To overcome the large lag of the plant and existence of meal disturbances,the ultra-local modeling and dynamic linearization technology are combined to obtain the ultralocal data model with disturbance.Based on this,a prediction formula is derived for multistep prediction.Then revising feedback and rolling optimization are combined with the prediction formula,which obtains a model free adaptive predictive control with disturbance compensation(DC-MFAPC)algorithm.Finally,effectiveness and security of the algorithm is verified through numerical simulation and comparison.4)A maximum likelihood parameter estimation and K nearest neighbor disturbance compensation based model free adaptive predictive control(ML-KDC-MFAPC)algorithm is designed.In order to improve the adaptiveness of the control algorithm and the ability to resist meal disturbances,based on DC-MFAPC,the maximum likelihood method from Bayesian perspective is used for parameter estimation,and disturbance compensation is improved with modified KNN algorithm.Finally,ML-KDC-MFAPC is obtained,and its superiority is verified by simulation and comparison. |
PDF Full Text Request