Development of machine learning based energy consumption forecasting model for commercial buildings기계학습 기반 상업용 건물의 에너지 사용량 예측 모형 개발

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 162
  • Download : 0
As the building sector accounts for a considerable portion of global energy consumption and greenhouse gas emissions, effective management of building energy is of great importance. In this regard, forecasting building energy consumption is essential to use and manage the energy efficiently. It enables faster and more accurate energy analysis and an effective energy utilization plan. Various models are developed using traditional statistical approaches such as ARIMA, ARIMAX and artificial intelligence approaches such as neural network and deep learning. In this thesis, energy consumption forecasting model for commercial reference buildings provided by the U.S. Department of Energy(DOE) is developed using Long Short-Term Memory(LSTM) network model which is a special kind of RNN model that is capable of learning the long-term dependencies. Forecasting models for three regions in the U.S. with different climates are considered, reflecting the location of the selected reference building. Through weather input variable selection process, the model with one weather information input which is outdoor temperature and the model with three weather information inputs which are outdoor temperature, relative humidity and solar radiation are separately developed. The proposed LSTM model is compared with typical Artificial Neural Network model and NARX Neural Network model. The accuracy of the model is evaluated by MBE and CvRMSE and it is confirmed that all tested models satisfy the acceptable error range proposed by ASHRAE guideline 14. Although error varies from the seasons, areas to the forecasting techniques, not always but over half cases, LSTM models fit better than other network models. The results also showed that it’s not always better to reflect more weather input variables as the more cases showed better performances when applying outdoor temperature only as a weather information input when compared with the cases applying three weather information inputs which are outdoor temperature, humidity and solar radiation. The reliability of the simplified model only applying outdoor temperature for weather input is verified.
Advisors
Chang, Seong Ju장성주
Description
한국과학기술원 :건설및환경공학과,
Publisher
한국과학기술원
Issue Date
2019
Identifier
325007
Language
eng
Description

학위논문(석사) - 한국과학기술원 : 건설및환경공학과, 2019.8,[iii, 34 p. :]

Keywords

Building energy; energy consumption forecasting▼amachine learning▼aartificial neural network▼along short-term memory; 건물 에너지; 에너지 사용량 예측▼a기계학습▼a인공신경망▼a장단기 메모리

URI
http://hdl.handle.net/10203/282887
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=875099&flag=dissertation
Appears in Collection
CE-Theses_Master(석사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0