類神經理論應用於數據機房散熱節能之最佳化設計

The Optimal Cooling Layout of Data Center by Artificial Neural Networks Theory

程彥蓁(Yen-Chen Cheng)

指導教授 ： 湯敬民

淡江大學/工學院/航空太空工程學系碩士班/碩士(2013年)

https://doi.org/10.6846/TKU.2013.00133

摘要

為了趨緩地球暖化的問題，國內各大企業開始實施節能減碳的運動，而IT數據機房是實施的重點之一，根據調查，大多數的IT數據機房，為加強散熱的效果，隨意擺放增加的散熱裝置，但熱區域依舊高溫不下，只是浪費更多的能源。為了避免能源的浪費，本研究將針對機櫃與空調的配置關係進行研究，尋找出最適當的排列方式，達到建立低耗能、高效能的數據機房。
本研究首先建立類神經系統所需的範例資料庫，利用FDS(Fire Dynamics Simulator)模擬數據機房散熱的情況，在空間中佈放虛擬之Thermocouple，獲取溫度數據建置成範例資料庫，放進類神經網路系統進行訓練，找出最佳化機櫃走道寬度，再將其結果回饋至FDS中驗證，了解機櫃與空調機之間複雜的影響關係，找到適當的空調機放置方式。
綜觀所有的研究結果可知，當數據機房內有完整的循環氣流時，數據機房會有比較好的散熱表現。適當的利用冷氣流下降及熱氣上升的原理，可以增加數據機房散熱的效果。相同的空調設定，會因為機櫃與冷氣位置擺放的方法不同，造成不同程度的散熱效果。

關鍵字

類神經系 ； FDS ； 數據機房 ； 散熱

並列摘要

In order to ease the global warming problem, there were lots of companies started the green revolution in Taiwan. Most of the improvements were done in IT data center. According to the investigation, most of the IT data center placed additional air cooling in random ways. It couldn’t cool down the hot zone but wasted the energy instead. In order to avoid energy waste, this research used Artificial Neural Networks and FDS(Fire Dynamics Simulator) to deduce a better way to place air conditioners and venting exits. Such that the data center can be equipped with better cooling and performance.
The cases data base composed of FDS simulation data was created for artificial neural networks. In order to provide the simulation data, FDS computed the temperature distribution in the data center. Procedures in the case data center then trained and optimized the networks. The final results were fed into FDS again to verify and analyze the temperature variation in the data center.
The results showed that the circulated air stream enhanced the cooling effect. To improve cooling effect and reduce the loading of cooling systems, the heated air should be extracted from the upper part of the room since the warmer air was observed to accumulate in the region. Finally, the cooling effect would vary with different space arrangement even with the same air conditioner settings.

並列關鍵字

Cooling ； FDS ； Data Center ； Artificial Neural Networks

參考文獻

[4] 許筱琪,”大樓建築自然通風效果模擬研究”,冷凍空調&能源科技,2011

[3] 韓選棠,徐嘉宏,汪孟欣,” 住宅建築利用熱緩衝空間對冷氣使用耗電量影響之 研究-以台大綠房子為例”,農業工程學報,第56卷第2期,2000

[5] G. Carrilho da Graca, Q. Chen, L.R. Glicksman, L.K. Norford,”Simulation of wind-driven centilative cooling systems for an apartment building in Beijing and Shanghai”, Energy and Buildings, Vol.34,2002

[7] C. E. Bash, C. D. Patel, R. K. Sharma , ”Dynamic Thermal Management of Air Cooled Data Centers “, Hewlett-Packard Laboratories, Palo Alto, CA

[8] U. Singh, A. K Singh, Parvez S, A. Sivasubramaniam, “CFD-Based Operational Thermal Efficiency Improvement of a Production Data Center,” Tata Consultancy Services Ltd., India