期刊
ENERGY
卷 128, 期 -, 页码 688-700出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.energy.2017.04.034
关键词
Predictive energy analytics; Climate-energy nexus; Dew point temperature; Statistical learning theory
资金
- NSF SEES [1555582]
- Div Of Chem, Bioeng, Env, & Transp Sys
- Directorate For Engineering [1555582] Funding Source: National Science Foundation
The link between climate and energy needed for space conditioning has been widely studied, but is still beset with uncertainties. The uncertainties are further compounded by the projected climate change impacts, which will likely increase the frequency of weather and climate extremes. Unlike previous approaches, we leveraged non-linear statistical learning methods to develop predictive models for residential and commercial electricity usage to understand the relationship between weather, climate and electric power consumption. Our analyses were implemented for the state of Florida, one of the most energy intensive states with geographic and demographic vulnerability to climatic change. Our results indicate that the method of Bayesian additive regression trees (BART) best captures the complex structure of the data. We conclude that mean dew point temperature is a more suitable predictor of climate sensitive load than the most widely used degree-days variables. Moreover, wind-speed and precipitation are key predictors of the climate sensitive portion of electricity consumption. We also find that the residential electricity consumption has more spatio-temporal heterogeneity and is more influenced by consumer behavior than the commercial sector. (C) 2017 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据