以深度學習分析智慧電表資料的加值應用－以用電不安全預警為例

DEEP LEARNING-BASED SMART METER DATA ANALYTICS FOR EARLY WARNING OF POSSIBLE ELECTRICAL FIRES

10.6652/JoCICHE.201904_31(2).0007

鄭安平(An-Ping Jeng)；王如觀(Ru-Guan Wang)；吳佰餘(Pai-Yu Wu)；謝春棋(Chuen-Chyi Hsieh)；戴荏國(Jen-Kuo Tai)；談家成(Jia-Cheng Tan)；周建成(Chien-Cheng Chou)

深度學習 ； 時間資料庫 ； 智慧電表大數據分析 ； 電氣火災 ； deep learning ； temporal database ； smart meter data analytics ； electrical fire

中國土木水利工程學刊

31卷2期（2019 / 04 / 01）

193 - 204

繁體中文

在節能減碳趨勢下，低電壓智慧電表逐漸普及。智慧電表收集資料除可用作耗電量預測促使節電，亦有其他加值應用。本研究探討用電不安全預警，以深度學習法，預測瞬間超高用電量與持續高用電量情形。英國智慧電表大數據為公開資料且品質佳，台灣智慧電表格式確定但尚在累積資料中，本研究比較兩地資料，構思一致方法與因地制宜參數，設計時間資料庫供演算法訓練，並評估英國用電不安全預警成效。台灣則以相同作法、本地資料，初探模型預警效能。待台灣智慧電表建置完善，套用本研究成果，預期能降低住戶電氣火災機率，提高智慧電表安裝意願。

As there is an increasing number of countries worldwide to start large-scale deployments of smart meters to promote energy conservation and carbon reduction, residents' willingness is a factor that cannot be ignored. This research proposes a new type of smart meter data analytics that can provide residents with early warning of possible electrical fires, in order to prevent disasters to increase their enrolment and adoption of smart meters. Additionally, since the smart meter open data sets from UK are relatively complete, and because smart meters in Taiwan are currently being deployed, this study first compares the data formats of smart meters from the two countries and designs a consistent, deep learning-based algorithm that can be utilized for both UK and Taiwan cases. A temporal database is created to serve as the training data source, and such early warning models are then developed and tested. The results show that once a sufficient number of electricity consumption records are available, the proposed approach can predict whether there will be any instantaneously or continuously abnormal electricity consumption events during the next several hours. The prediction accuracy for such unsafe electricity usage is above 70%. As for the Taiwan case, with appropriate parameters adjustment and customization efforts, it can be expected that the proposed approach can help Taiwan residents detect possible electrical fires by using their smart meter data sets in order to ensure their safety of life and property.

1. Azar, E.,Menassa, C. C.(2014).Framework to evaluate energy-saving potential from occupancy interventions in typical commercial buildings in the United States.Journal of Computing in Civil Engineering,28(1),63-78.
2. Barker, S.,Mishra, A.,Irwin, D.,Cecchet, E.,Shenoy, P.,Albrecht, J.(2012).Smart*: An open data set and tools for enabling research in sustainable homes.Proceedings of the 2012 Workshop on Data Mining Applications in Sustainability (SustKDD 2012),Beijing, China:
3. Chou, C. C.,Chiang, C. T.,Wu, P. Y.,Chu, C. P.,Lin, C. Y.(2017).Spatiotemporal analysis and visualization of power consumption data integrated with building information models for energy savings.Resources, Conservation and Recycling,123,219-229.
4. Commonwealth of Australia, “Electricity consumption benchmarks [Web page]”, https://data.gov.au/dataset/electricity-consumption-benchmarks, Access date: 2019/3/1 (2015).
5. Edwards, R. E.,New, J.,Parker, L. E.(2012).Predicting future hourly residential electrical consumption: A machine learning case study.Energy and Buildings,49,591-603.
6. Google, “Get Started with TensorFlow [Web page]”, https://www.tensorflow.org, Access date: 2019/3/1 (2019).
7. Hebrail, G. and Berard, A., “Individual household electric power consumption Data Set - UCI Machine Learning Repository [Web page]”, https://ccchou.page.link/naxz, Access date: 2019/3/1 (2012).
8. Kleiminger, W.,Beckel, C.,Staake, T.,Santini, S.(2013).Occupancy detection from electricity consumption data.Proceedings of the 5th ACM Workshop on Embedded Systems for Energy-Efficient Buildings,Roma, Italy:
9. Mauriello, S., “SmartMeter Moratorium [Web page]”, https://ccchou.page.link/82KB, Access date: 2019/3/1 (2012).
10. Murray, D.,Stankovic, L.,Stankovic, V.(2017).An electrical load measurements dataset of United Kingdom households from a two-year longitudinal study.Scientific Data,4,1-12.
11. Pecan Street, “One of a kind — We’re helping the world’s best minds solve the biggest water and energy challenges [Web page]”, https://www.pecanstreet.org, Access date: 2019/3/1 (2019).
12. Preferred Networks, “RNN language models [Web page]”, https://docs.chainer.org/en/stable/examples/ptb.html, Access date: 2019/3/1 (2015).
13. Rahman, A.,Srikumar, V.,Smith, A. D.(2018).Predicting electricity consumption for commercial and residential buildings using deep recurrent neural networks.Applied Energy,212,372-385.
14. Sangani, K.(2013).Sangani, K., “In the safety of our own homes,” Engineering & Technology, January, pp. 46-48 (2013)..
15. Shi, H.,Xu, M.,Li, R.(2018).Deep learning for household load forecasting - a novel pooling deep RNN.IEEE Transactions on Smart Grid,9(5),5271-5280.
16. Wang, Y.,Chen, Q.,Hong, T.,Kang, C.(2019).Review of smart meter data analytics: Applications, methodologies, and challenges.IEEE Transactions on Smart Grid,10(3),3125-3148.
17. 中央氣象局，「氣候統計 \ 月平均(1981-2010) \ 台灣各地 [Web page]」，https://www.cwb.gov.tw/V7/climate/monthlyMean/Taiwan_tx.htm, Access date: 2019/5/1 (2019)。
18. 王金墩，「台電智慧電網未來規劃與運作 – 台電綜合研究所 [Web page]」，https://ccchou.page.link/hi8E, Access date: 2019/3/1 (2018)。
19. 台大智活中心，「能源人物誌 [Web page] 」 ，http://energytofu.ntu.edu.tw/p_list.html, Access date: 2019/3/1 (2019)。
20. 林常平,陳貽評(2011)。從消費端看智慧電表問題。能源報導，2011(9)，8-10。
21. 消防署，「資訊公開 \ 統計資料 \ 建築電氣火災統計– 內政部消防署 [Web page]」，https://www.nfa.gov.tw, Access date: 2019/3/1 (2019)。
22. 能源局，「智慧電網總體規劃方案 – 經濟部能源局[Web page]」，https://ccchou.page.link/xmZX, Access date: 2019/3/1 (2017)。
23. 許志義,楊梓萱,陳志綸,柳育林(2017)。智慧揭露應用於電力服務之研究。臺灣能源期刊，4(3)，345-360。
24. 陳弘毅,紀人豪(2016).火災學.台中:鼎茂圖書.

1. 談家成,張智雄,周建成,吳佰餘,王如觀(2020)。以深度學習與建築資訊模型及虛擬實境技術探討室內聲音定位。中國土木水利工程學刊，32(5)，383-392。