Performance monitoring and prediction of a domestic split-type air conditioner

It has been very challenging to accurately monitor and predict the performance of air conditioners due to system operation complexity and huge number of already installed systems. This has made it cumbersome and somewhat difficult to accurately ascertain the effect of this technology on the grid, the amount of achievable energy savings, emission reduction and water usage savings upon replacing traditional space conditioning devices with this system, especially for countries like South Africa whose primary source of energy is coal. The accuracy of the existing system monitoring and prediction methods is low. This paper intends to develop four models of higher accuracy that monitor and predict the heating and cooling performance in terms of COP and energy of a domestic split-type air conditioner. These multiple linear regression models were built via data experimentally obtained for environmental, system thermal variation and human behavioural variation predictors. With a high correlation existing between the predictors and the various responses (correlation coefficient of models between 0.95 and 0.97), the models possessed a determination coefficient of between 0.90 and 0.94. Hence, the developed models have a higher accuracy in predicting system performance irrespective of the season of operation. It was also realised that just about 57.63% of total daily heating energy is used for daily cooling by the system.