Energy demand patterns in an office building: A case study in Kraków (Southern Poland)

Abstract

Office buildings play a significant role in shaping the current electricity demand and its trends. Their energy demand patterns impact the power system operation on a national and regional level. What is more, both (office buildings and the power system) are also simultaneously influenced by meteorological parameters. Considering the above, the aim of this paper is to analyze a three-year hourly energy demand time series recorded in a relatively large office building in Kraków (Southern Poland). This paper will fill a gap in the literature as there is a lack of evidence from Central European countries in the area of office buildings' energy demand and its relationship with meteorological parameters. The data was obtained from a local electricity provider whereas meteorological parameters came from weather station and satellite measurements. The analyses focused on determining the typical weekly and daily demand patterns as well as on investigating the correlation between meteorological parameters (wind speed, irradiation, humidity, and air temperature) and observed energy consumption. To estimate the correlation between investigated variables, a Pearson coefficient of correlation was used. For distinguishing typical load patterns, a k-means clustering method was applied. The relationship between meteorological parameters and load was also tested based on multiple linear regression analysis. The results indicated that energy demand had a relatively strong positive correlation with irradiation and with temperature and a negative one with humidity. The correlation with wind speed was not greater than 0.25. Dividing the data into three subsets shows that energy demand generally exhibits a stronger correlation with meteorological parameters on working days. Additionally, clustering analysis has shown that it is possible to distinguish three typical daily patterns of energy demand and meteorological parameters that correspond to a hot/warm day, cold days and days that are intermediary between those two. The regression analysis showed that meteorological parameters can explain/model a significant part of the load variability (up to 50%) although the quality of such models is relatively poor (in terms of mean absolute percentage error the best model exhibited a value of 16%). The results of this study can be used as a benchmark for similar office buildings that received the same level of sustainability certification, or in the future analysis of climate change impact on power demand.