Research On Intelligent Control Algorithm Of Human Complicated Blood Glucose

Keeping glucose in a normal range is important to make the organs work well, and it will have a bad influence on patients'health if the glucose is too high or too low. Considering there are many kinds of patients need the glucose control, the control method should not be the same. It is important to make different treatments for different patients. In this project, we introduce two different glucose control algorithms for patients with type I diabetes and intensive care patients:(1) Since type I diabetes patients has a lot of experience with the insulin dose, to take good use of the history data and conquer the weakness of continuous blood glucose monitoring system that the monitoring value may not be accurate, an algorithm includes more history data and statistic data is used. In this article, a method considering the difference between the effects of different periods of time history data on the current batch. A high order Run-to-Run method with forgetting factor is proposed. With using this new high order Run-to-Run method, the record of latest seven days can be used with different weights. During the simulation, the proposed method shows good result when test on ten different virtual patients. Though the patient's parameter has lots of differences, the method can get acceptable result for every virtual patient.(2) For patients with intensive care (ICU), even if there is no history of diabetes, stress hyperglycemia may occur, it will have bad effect on the treatment of the disease. Because the intensive care patients has no history of blood sugar control, and usually the condition is urgent, so we need to quickly identify the blood glucose model according to the patient's individual situation, to control the disease in time. According to the characteristics of patients in intensive care unit, this paper presents an modified cuckoo algorithm, compared to traditional particle swarm optimization algorithm when achieve the similar control result it takes less time on modeling. In the test of thirty virtual patients, it indicated that using modified cuckoo algorithm combined with model predictive control can keep the blood glucose level of intensive care patients in a stable range.