Search

This study aims to conduct data mining research on the time series energy consumption dataset of a small hotel. Earlier studies on data mining have demonstrated that cluster and association analysis had been commonly used methods today, while this has not yet been investigated under time series dimension. For that consequence, this article utilizes K-shape and Apriori algorithm coded by Python language to explore the time series energy consumption data of the small hotel in terms of the subentry and total energy consumption. Final results reveal that the energy consumption curve and association rules can effectively reflect the working characteristics of the small hotel. From the clustering results, the small hotel working feature and weather condition determine the time series energy consumption curve shape. As for association rules, there is a different chain relationship between the energy consumption of each subentry and total energy consumption of the small hotel, especially the consumption of heating gas and cooling, which mainly determines the changes in other energy consumption. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to study the energy usage pattern of office building all year round using time-series energy consumption data. This investigated pattern could uncover the energy utilization issue availing building energy efficiency implementation. Past researches principally focused on the total energy condition instead of time-series in terms of time cycle. This paper implements the innovative artificial intelligent algorithms to perform the data mining target via cluster analysis and association rule discovery between different types of energy. Official energy building models provide the studied database. The result shows that k-shape and apriori algorithm could successfully obtain the energy using pattern hidden dataset. There is significant dispensable energy wastage after working time with office building since the long leave course happening after 18:00. The main subentry energy determining total energy is different in disparate stages. Moreover, cooling energy primarily manipulates the total energy in most of time indicating more than 80% degree in terms of confidence level. Conclusion illustrates that this workflow could successfully detect power load profile features and find the unreasonable issues with energy using. Combining above cluster and association analysis outcome contributes to the energy adjustment in each period more precision. [ABSTRACT FROM AUTHOR]