Black-box modeling and adaptive control of hybrid thermoelectric refrigerator systems
Abstract
Vaccines carrier has been used to keep the temperature within 2–8°C. However, a
poorly functioning vaccines carrier may expose the vaccines to freezing temperatures.
Hybrid Thermoelectric Refrigerator (H-TER) systems are developed in order to
transport sensitive vaccines to hospitals at accurate controlled temperature. The
refrigerator use air-to-air and direct thermoelectric heat pumps. This work reports on
modeling and control studies carried out for H-TER systems that can control low
temperature accurately. Different type of material containers (aluminium and stainless
steel) is used and comparisons between them are analyzed. The systems are nonlinear
and exhibits varying model parameters and dead-time. The objective of the study is to
investigate control strategies that are based on non-priori plant knowledge and yet
allowing for continuous adaptations of the controller to changing system dynamics. In
fact, the various cooling load also causes a reduction of refrigerator efficiency including
the fluctuation of imposed current level due to electronic component performance and
varied operating condition of thermoelectric module on cold and hot ends against time.
Thus an adaptive control system is considered to handle the problems that are stipulated
above. A black box modeling approach is chosen since this is needed for the
implementation of adaptive controllers. The H-TER systems have been identified using
both Recursive Least Squares (RLS) and Recursive Extended Least Squares (RELS)
methods. Since RELS has shown to give biased estimates for filtered data and slow
convergence estimates for unfiltered data, RLS has been chosen for the model as its
give a better representation of the systems. A second order model of H-TER I and HTER
II systems are found to adequately represent the system as it give best fit of 0.0009
and 0.0007 respectively which made the fourth order to be insignificant for
implementation. Validation procedures using second order model for online estimation,
show that the model is indeed a good representation of the H-TER systems. On-Off and
PI controllers are the commonly used in thermoelectric system is applied to this system
as case studies. PI controller shows better performance over On-Off controller in term
of steady state error.