Experiment-based Reduced Order Model by MIM applied to empty beehives for temperature prediction

Beekeepers are facing environmental challenges that threaten their profession. Many companies have developed wireless devices to help beekeepers to keep track of their colonies and to reduce unnecessary trips by monitoring the weight and the temperature of the hive in real time. Contrary to weight data, it’s not clear yet what information can be retrieved from the temperature.

In this study, an algorithm to predict in-hive temperature is tested on empty beehives. This algorithm is based on a reduced order model (ROM) by the Modal Identification Method (MIM). The ROM equations are derived from local heat transfer equations and involve matrices that are unknown and have to be estimated by using an optimization algorithm. In this work, the MIM model learning phase is experiment-based, i.e. it relies only on temperature measurements provided by beehives equipped with several temperature sensors. The close environment of the hive is characterized using a set of sensors (temperature, irradiance, anemometer, etc.).

This kind of algorithm may be integrated in monitoring devices to send alert to the beekeeper before critical temperatures are reached in the hive. One key advantage of experiment-based models is to avoid having to specify hive geometry, material properties and boundary conditions (sun exposure, wind, air temperature, surrounding hives or trees or objects, type of ground…) which are all hive-specific and, for some of them, time-varying.

In this article, the temperature evolution for twelve days of an empty Dadant hive is presented and compared to the reduced order model outputs. The model is identified using the first eight days. The last three days are used for the comparison.

This work is a collaboration between PPRIME-UPR CNRS 3346 and CoActions-AltRD company whose research are funded by the European Better-B project (101081444-Better-B) in the frame of European Programme HORIZON-CL6-2022-BIODIV-02-two stage on resilient beekeeping (2023-2027).

Article

PDF : download