Design and Analysis of a Novel Split and Aggregated Transmission Control Protocol for Smart Metering Infrastructure

摘要

Utility companies (electricity, gas, and water suppliers), governments, andresearchers recognize an urgent need to deploy communication-based systems toautomate data collection from smart meters and sensors, known as Smart MeteringInfrastructure (SMI) or Automatic Meter Reading (AMR). A smart metering systemis envisaged to bring tremendous benefits to customers, utilities, andgovernments. The advantages include reducing peak demand for energy, supportingthe time-of-use concept for billing, enabling customers to make informeddecisions, and performing effective load management, to name a few.A key element in an SMI is communications between meters and utility servers.However, the mass deployment of metering devices in the grid calls for studyingthe scalability of communication protocols. SMI is characterized by thedeployment of a large number of small Internet Protocol (IP) devices sendingsmall packets at a low rate to a central server. Although the individualdevices generate data at a low rate, the collective traffic produced issignificant and is disruptive to network communication functionality. Thisresearch work focuses on the scalability of the transport layerfunctionalities. The TCP congestion control mechanism, in particular, would beineffective for the traffic of smart meters because a large volume of datacomes from a large number of individual sources. This situation makes the TCPcongestion control mechanism unable to lower the transmission rate even whencongestion occurs. The consequences are a high loss rate for metered data anddegraded throughput for competing traffic in the smart metering network.To enhance the performance of TCP in a smart metering infrastructure (SMI), weintroduce a novel TCP-based scheme, called Split- and Aggregated-TCP (SA-TCP).This scheme is based on the idea of upgrading intermediate devices in SMI(known in the industry as regional collectors) to offer the service ofaggregating the TCP connections. An SA-TCP aggregator collects data packetsfrom the smart meters of its region over separate TCP connections; then itreliably forwards the data over another TCP connection to the utility server.The proposed split and aggregated scheme provides a better response to trafficconditions and, most importantly, makes the TCP congestion control and flowcontrol mechanisms effective. Supported by extensive ns-2 simulations, we showthe effectiveness of the SA-TCP approach to mitigating the problems in terms ofthe throughput and packet loss rate performance metrics.A full mathematical model of SA-TCP is provided. The model is highly accurateand flexible in predicting the behaviour of the two stages, separately andcombined, of the SA-TCP scheme in terms of throughput, packet loss rate andend-to-end delay. Considering the two stages of the scheme, the modellingapproach uses Markovian models to represent smart meters in the first stage andSA-TCP aggregators in the second. Then, the approach studies the interaction ofsmart meters and SA-TCP aggregators with the network by means of standardqueuing models. The ns-2 simulations validate the math model results.A comprehensive performance analysis of the SA-TCP scheme is performed. Itstudies the impact of varying various parameters on the scheme, including theimpact of network link capacity, buffering capacity of those RCs that act asSA-TCP aggregators, propagation delay between the meters and the utilityserver, and finally, the number of SA-TCP aggregators. The performance resultsshow that adjusting those parameters makes it possible to further enhancecongestion control in SMI. Therefore, this thesis also formulates anoptimization model to achieve better TCP performance and ensures satisfactoryperformance results, such as a minimal loss rate and acceptable end-to-enddelay. The optimization model also considers minimizing the SA-TCP schemedeployment cost by balancing the number of SA-TCP aggregators and the linkbandwidth, while still satisfying performance requirements.