Load allocation improvement for chiller system in an institutional building using logistic regression

Chiller systems are an essential part of central air conditioning systems to provide cooling energy in buildings. This study discusses how to improve the load allocation of chillers to reduce the system electricity consumption. An analysis of comprehensive operating data from an existing system with 5 chillers of 2 different capacities identified 101 types of component combination. The variation of system coefficient of performance depended significantly on the chiller part load ratio, followed by the temperature of cooling water leaving the condensers and the type of component combination. Chiller sequencing took place for only 30.1% of the total operating time. There was 61.73% of the total operating time in underload conditions with an average part load ratio of 0.65. Implementing chiller sequencing at all times increased the average part load ratio to 0.78 with a 15.81% reduction in the electricity consumption. Based on the results of multi-nominal logistic regression, the underload conditions could be predicted by an increase in the temperature of supply chilled water and the temperature of cooling water entering the condensers. On the other hand, an increase in the temperature of returned chilled water and the temperature of cooling water leaving the condensers reflected a higher chance of overload conditions. Controlling system components with the trend of significant temperature variables was proposed to enhance chiller sequencing.