Parametric Identification of a New Model for Type 1 Diabetic Patients

PABLO S. RIVADENEIRA; JOSÉ L. GODOY; LUISA F. COLORADO

Cartagena

Conferencia; 2017 IEEE 3rd Colombian Conference on Automatic Control (CCAC); 2017

IEEE

Inthis paper, the identification of a new relevant control model for type 1diabetic patients is developed, which allows computing tools of the functionalinsulin therapy directly from its parameters. Also, this new model possesses equilibriumsaccording to the physiological behavior, i.e. i) without insulin, the glycemiaraises indefinitely, ii) with basal insulin, the glycemia is maintained at itsactual value (it does not go to the basal levels, as it is stated by thestandard models), and iii) with a carbohydrate input, the glycemia raises and remainsat a high value if no insulin is administered. For the identification stage,different structures of the model are analyzed with different identificationalgorithms in order to select the best structure and algorithm (which minimizesthe fit error). A set of data collected during 3 days from real and virtual patients(from the UVA/Padova simulator) was used to identify the model. The fitobtained was 20% y70% forreal and virtual patients, respectively. The conclusions of this study are i)the proposed model even its simplicity can be used to represent diabetic patientsand ii) this promotes its potential application to the so-called artificialpancreas in the design of control strategies, as model predictive control.