Energy consumption prediction of air-conditioning systems in buildings by selecting similar days based on combined weights

Abstract

Accurate modelling and prediction of energy consumption of the air conditioning system is crucial for improving decision making. A method for predicting the energy consumption of air-conditioning systems is proposed in this paper. Based on the same weather type (sunny, cloudy, overcast, or rainy) and day type (workdays or holidays), the similarity errors using the combined weight method and the baseline errors of similar working conditions are calculated with this method. These conditions include outdoor temperature and lighting and plug power, then, similar days are determined within a certain similar error range. In addition, the air-conditioning energy consumption in these similar days is regarded as that in the predicted days. The similarity errors in selecting similar days are acquired by efficiently combining subjective weights, objective entropy weights, and correlation coefficients. To verify the accuracies of the predicted energy consumption using similar days method based on combined weights, a simulation was performed by eQUEST. According to the simulation example of measured data in an office building, it is proved that the proposed prediction method with high forecast accuracy can select similar days with a high degree of similarity under non-catastrophic weather conditions, and offer promise for wider engineering application.