New algorithms for daylight harvesting in a private office

A daylight harvesting system can produce almost 60% energy saving by using daylight to satisfy illumination requirements. In this paper we study the problem of daylight harvesting for an indoor office which has adjustable electric lights and blinds. There are two main problems with such a system namely; a) photo sensor should ideally be placed on the work desk to measure illumination. However such a sensor can be accidently covered by paper or occluded by user leading inaccurate measurements b) changing daylight distribution inside the room due to movement of sun and window blinds. New data fusion algorithms are proposed in this paper that employ machine learning and radiosity theory to compute the electric light and daylight component (hence total illuminance) on the work plane using ceiling mounted sensor. Experimental results from a real test bed are provided at the end to highlight the performance of each algorithm.