Towards a Smart Operation: Novel Grey-box Modelling of Ultra-low Temperature Freezing Chambers

This study develops a novel grey-box approach for modelling the heat dynamics of the ULT freezing chamber. The modelling approach is based on stochastic differential equations and fed only with costless data measured by embedded temperature sensors in the particular freezer under study. The unknown parameters are estimated using the maximum likelihood method based on data from regular operation periods. A novel model that describes the nonlinear heat transfer between the local evaporator and chamber is proposed. This nonlinear model is able to compensate for the biased measurements from the embedded temperature probe and adapt to changes in the local cooling intensity. Two freezers with different operation patterns are modelled. The results prove that the established models perform well in predicting chamber temperature (RMSE < 0.19°C), which is the backbone of reliable applications. Despite compromised physical interpretability, the developed modelling approach eases the modelling procedure and is expected to greatly promote the potential of practical implementations of grey-box models in future intelligent surveillance and operation of ULT freezers.