利用嵌入式系統建構分散式私有雲─以電力品質辨識為例

Abstract

本研究利用嵌入式系統結合雲端運算中私有雲伺服器的概念，研發了一套具有高擴充性的分散式運算私有雲分析系統(Distributed Private Cloud Server, DPCS)，所謂的高擴充性為雲端運算的一大特點，即不須重新啟動整套系統，即可動態的加入或移除網路內的運算成員，也正因為如此，本研究所提出的系統不會因為系統中某一兩個運算單元損毀而導致整套系統崩潰。 此套系統可設置於無線感測器網路架構當中位於前端感測區域內的閘道器上面，透過路由器的鏈結，閘道器在收集到大量的原始感測資料時，能夠透過本研究所提出的系統在感測區域前端就能進行資料分析，只需將分析好的結果回傳至後端監控平台，達到資料探勘的精神。在硬體方面，本研究所採用來當作伺服器的運算單元為嵌入式系統，嵌入式系統在設置的成本上相較於大型公司所採用大型運算伺服器的成本低了許多，在分散式私有雲程式方面，本研究提出一個以系統CPU使用率來當作指標，結合機率公式去將大量的資料量分散送給不同的運算單元做分析，達到負載平衡的作用。 電力品質的汙染問題逐年受到重視，為此，本研究以電力資料為例，在本研究提出的分散式運算分析系統上加入一些分析電力品質的方法，用以分析的電力資料，目前本研究根據IEEE-1159的標準已完成三種關於電壓變動事件的分析，此三種電壓變動事件分別為: 電壓突升、電壓突降及電壓中斷三種電壓變動事件，在資料分析的過程中若發生電壓變動事件，本研究提出的分析系統將會即時回報結果給後端平台，讓使用者或管理者在第一時間就能得知電力品質問題發生的時間點。 實驗驗證本研究所提出系統具備高擴充性、資料探勘及負載平衡的功能，未來此套私有雲分散式運算分析系統將不只侷限於電力品質問題的分析上，此分析系統可望應用在各種不同類型;資料分析上面，讓分散式私有雲伺服器的應用範疇更為廣泛，造福人群。

In this study, an embedded system-based distributed private cloud server (DPCS) was designed and implemented. The DPCS system has high scalability which is a feature of cloud computing. High scalability means that users do not need to reset the entire system. The computing nodes of a high scalability system can join or leave the network without reconfiguration. Because of the high scalability, the breakdown of a few computing units will not lead to the crash of the proposed system. The proposed system can be installed on the gateway in a wireless sensor network deployed in a sensing area. When a large amount of sensing data collected by the gateway, the proposed system can analyze the sensing data immediately and then send the analysis results to the backend server to achieve the goal of data mining. Compared with the servers of larger companies, the cost of embedded systems is much lower. In this study, combed with the probability method, a distributed data packet method based on the CPU usage was proposed. The proposed method allows the DPCS to distribute the large amount of sensing data to achieve the goal of load balancing. The issue of power quality has been becoming more important recently. Taking power quality monitoring as an example, a few analysis methods were used by the DPCS system to analyze the electric data. According to the IEEE-1159 standard, the proposed system could analyze three types of voltage variation: voltage swell, voltage sag and voltage interruption. If a voltage variation event is detected by the DPCS, the DPCS will send a message to the backend server, and the administrators or users can get this information immediately. The research experiments verify that the proposed system has the merit of high scalability and is capable of data mining and load balancing. In the future, the DPCS can used to analyze not only power quality but also more kinds of sensing data to improve the quality of the lives of many.