Prediction of the cop of existing rooftop units using artificial neural networks and minimum number of sensors

This paper proposes a new approach for evaluating the Coefficient of Performance (COP) of existing rooftop units, using the General Regression Neural Networks. This approach reduces the installation cost of monitoring equipment since only a minimum number of sensors is needed, and it also reduces the costs for re-calibration or replacement of sensors during the operation. The new approach was developed and tested using measurements taken on two existing rooftop units in Montreal, Canada.