Critically ill patients are known to experience stress-induced hyperglycemia. Inhibiting the physiological response to increased glycaemic levels in these patients are factors such as increased insulin resistance, increased dextrose input, absolute or relative insulin deficiency, and drug therapy. Although hyperglycemia can be a marker for severity of illness, it can also worsen outcomes, leading to an increased risk of further complications. Recent studies have shown that tight control can reduce mortality up to 43%. Metabolic modelling has been used to study physiological behaviour and/or to control glycaemia for a long time and many successful approximate system models have been developed. Due to the malfunction of medical equipments, clinical measurements obtained usually come with noise. In addition, the few such systems currently available can have errors in excess of 20-30%. Therefore, to fully simulate the clinical data, the system model also needs to couple with a successful noise model. This research has developed a new noise model that better fits the current available statistical description of the noise profile and therefore can be applied to achieve better simulation results. The research also designed a filter algorithm that is capable of reducing the sensor measurement error down to an acceptable value. Achieving such a goal is a significant step towards fully automated adaptive control of hyperglycaemia in critically ill patients and would therefore reduce mortality.