預測性推論之預主動式情境感知節能系統

Anticipatory Reasoning for a Proactive Context-aware Energy Saving System

曾毅修(I-Show Tseng)

指導教授 ： 傅立成

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.10063

摘要

由於能源的減少，節能已經成為近年來的一個熱門話題。大多數的家庭節能研究專注在建立一個反應式的節能系統，只考量當下的情境來做決定。他們忽略了預測未來情境的節能潛力。除此之外，因為無法預先做出反應，反應式的節能系統也會造成使用者舒適度的損失。大部分專注在主動式節能的研究主要是利用占用模式來進行預測。由於缺乏活動的資訊，他們無法提供細膩的控制，可能會導致不全面的節能以及使用者舒適度下降。而且他們也很少考慮到人行為的不確定性，而只使用學習到的模式來進行預測。另一種類的研究會利用一些與電器或是活動相關的模式，例如電器使用的順序性模式，或是一個活動發生時間及長度的週期性模式來進行預測。然而，他們通常只會在同一時間使用其中一種模式，但由於人性的複雜性只使用其中一種模式是不足以達到良好的預測準確性。
本研究的貢獻有以下三點： 第一，我們會就活動的時間延遲這方面來考量人行為的不確定性以更佳的描述人的行為。第二，我們將善用混合模式的益處同時利用順序性模式和週期性模式來增進預測的準確性，讓我們能夠提供更加能被接受的主動式節能服務並且降低錯誤服務的代價。最後但同樣重要的，我們專注在活動預測來提供細膩的控制以達到更為全面性的節能以及提升使用者舒適度。

關鍵字

節能系統 ； 活動預測 ； 預主動式 ； 情境感知 ； 使用者舒適度

並列摘要

Because of the decrease of energy source, energy saving has become a popular issue in recent years. Most of the prior works focus on building a reactive energy saving system to achieve energy saving, of which the decision is made only based on the current contexts. They neglect the potential of more energy saving if we can anticipate the future contexts. Besides, reactive energy saving systems could also cause comfort loss of the inhabitants because they are not able to react in advance. Those works that focus on proactive energy saving mainly use occupancy pattern for prediction. Lacking the activity information, they are not able to provide more fine-grained control, which could result in incomplete energy saving or decreasing of user comfort. What’s more, they seldom consider the uncertainty of human behavior and just use the learnt pattern for prediction. There is another class of results exploiting some human behavior patterns relevant to appliance usage or daily activities, such as sequential pattern of using appliances or periodical pattern of start time and duration of activity, for prediction. However, they usually use either one of the two patterns at a time, which is not enough to achieve good prediction accuracy due to the complexity of human nature.
The main contribution of the thesis is three-fold. Firstly, we will consider the uncertainty of human behavior in terms of delay of an activity, which could depict the human behavior better. Secondly, we will exploit the hybrid patterns, namely, periodical pattern and sequential pattern, for more accurate prediction, which enabling us to provide more acceptable proactive energy saving services and to decrease the cost of false services. Last but not least, we focus on the activity prediction to provide fine-grained control for a more thorough energy saving and increasing user comfort.

並列關鍵字

Energy Saving System ； Activity Prediction ； Proactivity ； Context-aware ； User Comfort

參考文獻

[1] "COM(2008) 30 final/, COMMUNICATION FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT, THE COUNCIL, THE EUROPEAN ECONOMIC AND SOCIAL COMMITTEE AND THE COMMITTEE OF THE REGIONS-20 20 by 2020-Europe's climate change opportunity."

[2] F. Unander, I. Ettestol, M. Ting, and L. Schipper, "Residential energy use: an international perspective on long-term trends in Denmark, Norway and Sweden," Energy Policy, vol. 32, pp. 1395-1404, 8// 2004.

[3] P. Holtedahl and F. L. Joutz, "Residential electricity demand in Taiwan," Energy Economics, vol. 26, pp. 201-224, 3// 2004.

[4] U.S. Energy Information Administration. Available: http://www.eia.gov/todayinenergy/detail.cfm?id=10271&src=%E2%80%B9%20Consumption%20%20%20%20%20%20Residential%20Energy%20Consumption%20Survey%20(RECS)-f1

[5] E. P. B. E. S. Energy, "Forecasting Division. Canada’s energy outlook, 1996-2020," Natural Resources Canada, 1997.