Prediction Model on Room Temperature Side Effect due to FastADR Aggregation for a Cluster of Building Air-Conditioning Facilities

In this paper, we have constructed an auto regressive and neural network combined type prediction model for responsive change in the room temperature trend due to the fast automated demand response (FastADR) power limitation of office building air-conditioning facilities. We defined the average of differences between room temperature and set point of each indoor unit for the entire building as a FastADR side effect index for the building. Prediction experiments using an actual office building showed that the root mean square prediction error of our model was 0.23 degrees C for 5 min after the FastADR. This prediction ability is considered sufficient to estimate the side effect of FastADR power limitation.