Energy-Efficient Parameter Adaptation and Prediction Algorithms for the Estimation of Temperature Development Inside a Food Container

Abstract

This paper presents the development and implementation of energy-efficient parameter adaptation for a grey-box model representing the temperature profile in spatial points of the interior of a refrigerated container with the aim to improving the logistics of perishable goods. A mixed linear / non-linear singe-input-single-output grey-box model was selected for accurate prediction of the temperature behavior of the loaded food products. The algorithms were specially modified to reduce the matrix dimensions, implemented in Matlab, and applied to experimental data for validation. Apart from being highly accurate, the predictions comply with the desired figures of merit for the implementation in wireless sensor nodes, such as high robustness against quantization and environmental noise. The OSGi framework, which allows for easy update of software bundles, was selected as basis of the software implementation on the iMote2 as sensor network platform. Performance measurements have shown that this method provides a fast and accurate prediction with high energy efficiency.