智慧電網利用安裝在各用電戶的智慧電表收集用電戶的用電資料並進行分
析，使得電力公司可即時監控各地區的用電量並且進行效益的最佳化，另外可
讓用電戶可藉由用電量的分析來調整各時段用電量，以達到節約電費的最佳
化。建立智慧電網需建構先進讀表基礎建設（Advanced Metering Infrastructure,
簡稱AMI）的通訊網路；AMI 通訊網路的建構包含了智慧電表與集中器的佈置
以及相關通訊技術的選擇，其目標是讓智慧電表的用電相關資訊能夠先傳到集
中器，經由集中器透過通訊網路傳回電表資料管理系統。在這個計畫考量的環
境，智慧電表的資訊是經由其他智慧電表轉傳的方式透過RF-Mesh 無線網路傳
至集中器。由於集中器的成本很高，建構AMI 通訊網路須使用最少集中器使得
每個智慧電表都有至少一條轉傳的路徑到某集中器；因此，我們需要預測是否
存在至少一條轉傳的路徑在給定一組集中器位置及智慧電表位置間，即為本計
畫的最終目標；然而，在本文中，我們所要探討的是預測兩電表間存在路徑的
機率之實驗與探討。

Smart Grids (SGs) analyze the information collected by smart meters installed at
consumers. With SGs, the electric power company can maximize the profit by realtime
monitoring energy consumption of consumers and the consumer can minimize
the cost by arranging energy consumption into suitable time slots. Advanced
Metering Infrastructure (AMI), which is composed of smart meters, concentrators,
and communication technologies, is the foundation of SGs. The information
collected by smart meters can be first transmitted to concentrators for aggregation
via AMI mesh networks (e.q., RF mesh networks), and then trasmitted to Meter Data
Management Systems (MDMS) via wired or wireless backhauls, where a smart meter
transmits the collected information to the concentrator in a muli-hop manner in this
project. Due to the high cost of a concentrator, we need to minimize the number of
concentrators deployed in AMI mesh networks such that each smart meter has at
least one path to some concentrator. Thus, it is required to predict whether there exists
at least one path between an arbitrary pair of locations where the smart meter and the
concentrator reside respectively, which is the goal of this project. In this article, we
need to predict whether there exists a link between the meters.

[1] Mohassel, Ramyar Rashed, et al. "A survey on advanced metering infrastructure." International Journal of Electrical Power & Energy Systems 63 (2014): 473-484.
[2] Rolim, Guilherme, et al. "Modelling the Data Aggregator Positioning Problem in Smart Grids." Computer and Information Technology; Ubiquitous Computing and Communications; Dependable, Autonomic and Secure Computing; Pervasive Intelligence and Computing (CIT/IUCC/DASC/PICOM), 2015 IEEE International Conference on. IEEE, 2015.
[3] MacKay, David JC. "Chapter 20. An Example Inference Task: Clustering" Information theory, inference and learning algorithms. Cambridge university press, 2003.
[4] Popescu, I., et al. "Neural networks applications for the prediction of propagation path loss in urban environments." Vehicular Technology Conference, 2001. VTC 2001 Spring. IEEE VTS 53rd. Vol. 1. IEEE, 2001.
[5] Wölfle, G., and F. M. Landstorfer. "Field strength prediction in indoor environments with neural networks." Vehicular Technology Conference, 1997, IEEE 47th. Vol. 1. IEEE, 1997.
[6] Piacentini, Mauro, and Francesco Rinaldi. "Path loss prediction in urban environment using learning machines and dimensionality reduction techniques."Computational Management Science 8.4 (2011): 371-385.
[7] McLeod, Adam, Scott Bai, and Joseph Meyer. "Graphical Processing Unit accelerated radio path-loss estimation with neural networks." The Journal of Defense Modeling and Simulation: Applications, Methodology, Technology 10.2 (2013): 117-130.
[8] http://tgos.nat.gov.tw/tgos/web/tgos_home.aspx
[9] Ghosh, Joy, et al. "On a routing problem within probabilistic graphs and its application to intermittently connected networks." INFOCOM 2007. 26th IEEE International Conference on Computer Communications. IEEE. IEEE, 2007.