ETRI-Knowledge Sharing Plaform



논문 검색
구분 SCI
연도 ~ 키워드


학술대회 Sequence-to-Sequence model for Building Energy Consumption Prediction
Cited 7 time in scopus Download 1 time Share share facebook twitter linkedin kakaostory
김말희, 전종암, 김내수, 송유진, 표철식
International Conference on Information and Communication Technology Convergence (ICTC) 2018, pp.1243-1245
Gartner predicts that by 2022, 25% of utilities globally will use AI-augmented digital customer service agents to interact with customers' virtual personal assistants and home IoT (Internet of Things). Future power grids' important feature is the ability to predict the energy consumption over a different range of time spans. Temporal energy consumption prediction enables building managers to plan out the energy provision over time. Prediction makes it possible to shift energy use to off-peak periods, and makes more profitable energy purchase plans. But, building energy consumption prediction is a complex task because of many affecting factors, such as climate change trend, occupants' behaviors, and characteristics of thermal systems. In this paper, we developed a deep running model called Sequence-to-Sequence (seq2seq) model for time series prediction of energy consumption. The evaluation used the actual sensed data for three months.
KSP 제안 키워드
Affecting factors, Building energy consumption, Change trend, Climate Change, Customer Service, Energy Consumption Prediction, Energy use, Home IoT, Internet of thing(IoT), Over time, Sequence-To-sequence model