PRIME 媒體接取協定於智慧型電網應用之效能評估與改善

Abstract

近年來,由於電力系統的老舊與缺乏更新改善,漸漸的無法滿足現今社會需求,因此許多國家提出智慧型電網的概念來改善傳統電網的問題。智慧型電網整合了資訊、通訊技術與現有的電力系統,提供一個更完善可靠的供電系統;其中一項應用為住家電表與變電所之間的低壓電力通訊網路。在低壓電力通訊網路中,目前已有許多標準被提出,其中 PRIME 是一開放標準且目前在歐洲被廣泛使用。它提供了一套自動電表管理系統的通訊架構,主要針對在窄頻電力線通訊上實體層與媒體接取層的設計,不僅可應用於現階段的電表資訊網路上,未來也朝向智慧型電網應用發展。由於目前缺乏一套針對電力線在智慧型電網上應用的完善模擬平台,因此我們將 PRIME 所規範的實體層與媒體接取層實作於 NS-2 上,並依據我們所建立的模擬平台分析 PRIME 在電表通訊網路上的效能。根據模擬結果,我們發現了 PRIME 在通道競爭機制設計上的缺點;首先,由於複雜的網路建立流程造成網路建立所需時間過長,此外封包迴避碰撞機制無法有效避免控制封包於傳送過程中發生碰撞。其次,由於 PRIME 定義的競爭機制需花費較長時間,對於週期性的競爭機制較易發生封包碰撞且容易使得封包在競爭過程中被丟棄。在本篇論文中,我們首先提出「隨機偵測」來改善 PRIME 競爭機制的缺點,其方法主要是藉由隨機分散偵測時間來避免封包碰撞的機率。然而根據模擬結果顯示,此方法並未完全有效改善封包遺失的情形,因此我們利用電力線有線的特性使節點可以容忍偵測到通道忙碌,並藉由增加偵測通道的次數提高通道使用率,改善隨機偵測的方法。根據模擬結果顯示,改善後的方法可以有效減少一半的網路建立時間。此外在網路擁塞的情況下,相較於原來 PRIME 所設計的方法,此方法對於封包遺失率和延遲時間的改善可達 95%與 42%。

In recent years, the smart grid is proposed to address the traditional power grid problems. One of smart grid applications is the last-mile communication between utilities and customers in low-voltage distribution network. In last-mile communication systems, PRIME is an open standard that provides a telecommunication architecture for automatic metering management and it is supported by most of the European countries now. PRIME defines the physical layer and MAC layer of narrowband power line communication in low-voltage electricity grid. However, there is a lack of a full simulation platform for the power line communication in smart grid now. Therefore, we implement the power line physical layer and PRIME MAC layer in NS-2. Based on the simulation platform, we identify some problems in PRIME from the simulation result. One is the prolonged time for network formation due to the short backoff time and complex connection setup procedure. Another is the contention mechanism that causes the packets to be dropped during the transmission since the channel access mechanism wastes a lot of time. To address these problems, we propose a random sensing method to reduce the network formation time and packet loss rate by separating the sensing period to avoid collisions. However, initial simulation results show that this method is unable to solve all problems in PRIME. Therefore, we improve the random sensing method to further enhance the random sensing method performance by tolerating more occurrences of sensing a busy channel for enhancing the channel usage. Simulation results show that the improved random sensing method can reduce half of the network formation time, and it can improve the packet loss rate up to 95% and delay time up to 42% compared to original PRIME MAC when network is saturated.